Into the office again this morning to look at web server loads. Yes, gradually the new
server was getting some load, up to a load average of 2 on occasion. And the old server was
still round 20! What's causing that? Why
can't Apache log the web
server name?
While waiting for the web traffic to settle, moved on to the question of mail. Moved the
existing mail queue manually to eureka, and read:
Date: Sun, 29 Sep 2019 13:46:45 GMT
From: Mail Delivery Subsystem <MAILER-DAEMON>
Subject: Warning: could not send message for past 4 hours
----- Transcript of session follows -----
<grog@lemis.com>... Deferred: Operation timed out with mx1.lemis.com.
Warning: message still undelivered after 4 hours
Will keep trying until message is 5 days old
And indeed, the message was still there, a couple of days later. mx1 is
oldwww, and it's certainly not down. Tried
A bit of searching: yes, by default they block. I need to open a ticket and ask them to
unblock. OK, off to the web site to discover that I have to specify which server. Two
tickets where one would have done!
Quickly got a response telling me that the account team would look at it, and please don't
send duplicate tickets. OK, tell me how. Please reassign the ticket to be against their
web site limitations. New response: please don't submit duplicate
tickets. Sigh.
Shortly later got a question: what services do you provide, how much email per
day and per month, and what's the purpose of the mail? Clearly I answered to their
satisfaction, and I got a response:
We have removed the default SMTP block on your account. Please re-start your instances via https://my.vultr.com, for the change to take effect (re-starting the server itself _will_not_work).
Aaargh! Reboot! This isn't Microsoft. ffm has been up for 604 days. I don't want
to break that. And I don't want to interrupt web service. That really could have been done
better. For the time being I'm considering alternatives. I've missed the end of the month
deadline, so oldwww is with me until the end of the month. Maybe I'll be able to
bear rebooting by then.
In passing, why do these error messages use
the GMT time zone? It should
be UTC.
Apart from that, there seem to be issues with the name server. My understanding is that my
current config should serve requests from machines on the allow-transfer list,
but oldwww didn't want to know anything about www. Not a primary concern at
the moment, since I will be running a name server on www there too, and in the
meantime I can use the Vulture name servers.
It's the beginning of the month, time to finalize last
month's diary. Part of that involves passing the page through two different
validators at validator.w3.org: this
one for the HTML version and this for the RSS version. Most of the
errors are in copied markup, but this one baffled me:
Finally! There was a ctrl-A character at the end of the line. That explains the
obscure reference to non SGML character number 1. And once you know it, you can look
at the first of the messages from the HTML validator (but not the second) and see an
underscore character at the end of the line. And the XML message points at the r
in port, probably because it doesn't understand tabs.
We've been letting Nikolai run free while
walking for some time now, and things work well. But Yvonne didn't trust Leonid to do the same—until
today. Shortly before returning home (this photo is almost at the gate), she let him free.
At least here he behaved just as well:
Email from eBay today: do my grocery shopping
online and get it delivered to my door.
That's one of the borderline areas I considered in my 2014 article The future of the Internet, and most recently
here. How does it work? It seems that they're doing a deal with Coles, and that the delivery really is free, up to your
kitchen bench. There's a URL, https://www.ebay.com.au/help/buying/grocery-shopping-ebay/grocery-shopping-ebay?id=4873, but I doubt that it's very durable. The most important thing is that it's only in
selected metropolitan areas, the closest about 130 km away. They have my delivery address,
so they could easily have established that I'm not a target customer.
But where does eBay come into the equation? You'd think that Coles would do it themselves.
It seems that they do: here's another fragile
URL offering what appear to be different terms and conditions, and, in my case,
deciding that I live
in Pagewood, NSW. But
when I changed it to Dereel, they simply
modified their store to the one
in Sebastopol, which is the
correct store. There's no mention of any geographical restriction on deliveries, but I need
to spend at least $150, and it's only valid for this month.
PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND
57141 root 1 103 0 95668K 524K CPU1 1 117.5H 100.00% perl
Now that's something different. This is a 2 processor system, so a load average of 1.3 is
perfectly acceptable. But one of the processors was maxed out with a perl script,
something that I hadn't noticed in the mess of httpd processes. What is it?
Webmin! That's something that Chris Bahlo
uses (or used) to administer the site. Clearly something has gone wrong. But to think that
that has been accounting for 50% of the CPU load for the last few weeks!
And then another piece in the puzzle:
Date: Wed, 2 Oct 2019 12:23:17 +0000 (UTC)
From: World Wide Web Owner <www@www.lemis.com>
Subject: FAILURE: /grog/Thisshouldbemoving-in-1.jpeg.Isitmissing <- http://mail.lemis.com/grog/diary-jul1997.php
The “Thisshouldbe...” messages are an indication that the system didn't update the page
quickly enough, a result of the system overload. But look at that
URL: http://mail.lemis.com/grog/diary-jul1997.php. Where did they get that
from? That explains why the old system is still serving web pages. High time to get all
these spurious system names to redirect to the canonical name.
A while back I discussed on IRC whether the canonical name for my web site should
be lemis.com or www.lemis.com. The consensus was that lemis.com is
preferable. But that allows problems like this to occur. www.lemis.com it is.
Independently of this, it looks as if I should set
up HTTPS, which seems to have become the
only way to talk to web sites any more. Here the issue is a certificate. People have
recommended Let's encrypt, which led me on
to certbot.
Still to do:
Set up mail, which may involve migration to yet another Vulture.
Set up a redirect page for most of the names of the system.
Off to take the dogs for a walk, this time by myself. Today Nikolai came up as usual to have his chain put round his
neck, but I didn't put it on: lately I've been leaving him off the lead from the time we
leave the house, and it has worked well. But today he didn't want to leave the front of the
house. Tried to get him to come, but he demonstratively turned around and wanted to go back
into the house.
What's up? Is he sick? Took Leonid for a
brief walk, then back to look at Nikolai. How do you tell if a dog is sick? There were no
obvious symptoms. Cold, moist nose? Yes. But not as cold as Leonid's: 31° compared to
25°, as measured with a general-purpose infrared thermometer.
OK, should go by itself. But in the course of the day I discovered that he doesn't want to
go out the front of the house any more; normally he wants to go there several times a day.
What's up? Sick? Did something out the front of the house scare him? After a bit of
consideration, we're toying with the idea that he might feel rejected because he no longer
gets a leash. Could it be that what we thought was good for him gave him a feeling of
rejection? To be observed.
And the difference in nose temperature? Leonid had just come in from outside, while Nikolai
had been inside all the time. That, too, we should observe.
OK, as I noted at the time, they're intended to be grilled, not “air fried”, though I had my
doubts that that was the problem. Indeed. In the grill they not only burst, but also
charred:
„Schinkengriller“ is a German construction meaning “Ham grillers”. But these are clearly
bangers: „Schinkenplatzer“. A pity: they don't taste at all bad.
The weather was nice again today, so down Kleins Road with the dogs to walk in the area
where we went 5 years ago, from the “Eagle's Nest” (corner of Swamp Road and Kleins Road) to
the “Big Oak” (corner of Kleins Road and Swanson's Road).
The Acacia paradoxa are in full
bloom:
Niko (on a leash) was his usual self, though he did have a
little diarrhoea, so the cause of his
issues yesterday are still not clear. But at least he's livelier.
We've had goldfish in the garden for
over 10
years now. When we moved here from Kleins Road we kept them in a tub for a while
before putting them in the trough round the verandah, where they flourished.
Until about a month ago. After one particularly cold night Yvonne couldn't find any more, and thought that some bird
(cormorant?) had eaten them. That
sounded unlikely. I thought that they might have gone to the bottom of the water, where
it's warmer, and later I found one swimming around.
But that was the only time, and it really seems that they're gone. As if to prove the
point, the place now has lots of frogs, suggesting that there are no fish to eat the
tadpoles.
But how could they disappear? It's really hard to believe that a bird could extricate them
from their hiding places, and if they had just died we would see the remains. Some aquatic
predator?
Problems again today: my
remote Squid proxy refused
connections.
Why? Took a quick look. Not much activity, since it's just for me, but there was no
evidence of any problems. Nor of any connection attempts, even when I tried again.
Bingo! It's running on oldwww, but the configuration in my browser points
to www.lemis.com, now a different machine. Time to add a name squid.lemis.com.
Yvonne in today with a problem: she couldn't upload her
photos to the web server.
Oh. I had forgotten that. She doesn't use the server for anything else, but she does have
her photos there. And for that, the way I have it set up, she has her own account
and ssh credentials. Or at least, she had.
I had only set up accounts for Chris Bahlo and myself on the new machine. Not much work to
set it up for her, but just another indication of how easy it is to forget little details.
Into town for a number of things today. First stopped off at UFS: the infrared ear thermometer that I bought (but
didn't document) on 7 August 2019 seemed to work acceptably when I
bought it, but now it shows pseudo-random temperatures between 34° and 35.5°, clearly
incorrect.
Problem: I had mislaid the receipt. I had the credit card statement, but that was all. OK,
ask them to show me what I'm doing wrong (which, in fact, wasn't beyond the bounds of
possibility). Spoke to Keylin (if I remember the spelling correctly), who looked as if she
had never seen a thing like that before. She certainly couldn't tell me how to use it (not
even “stick
it in your ear”, which used to be an insulting suggestion).
After a bit of discussion with a pharmacist who was too polite to approach me, she suggested
calling the manufacturer. OK, why not? She was soon back: the whole company in on a
training course today and Monday, and they won't be back until Tuesday. She, too, found
that a little unusual.
OK, what's the alternative? The thermometer had cost $40, about as much as I was prepared
to pay. The next cheapest one cost round $80, and I could go up to $180 if I wanted. No,
not worth the trouble. Conventional in-mouth thermometers cost $7, and the number of times
I use one (the last one died of a flat battery) doesn't make it worth spending any more
money. Please refund.
That was easier than I thought. She didn't even ask for the receipt, just my membership
card. All was stored in their computer, which, however, required massive paperwork on her
part and two signatures from me before I got my refund.
On to ALDI, where I returned a smoke oven that
Yvonne had picked up last week. I had already tried what
appears to be the same device nine years ago, but last week I couldn't remember why we returned it. Lack of cold-smoking, maybe? I
have a solution for that. But no, the real issue is that it wasn't much good for
hot-smoking either, because it didn't have a thermostat and the flame couldn't be regulated
sufficiently.
While there, picked up some things that Yvonne had forgotten on Wednesday, and saw this
marvel:
Two portions of nasi goreng for $5!
What's in there? Mainly rice (400 g, corresponding to 150 g of raw rice, about $0.15
worth), an egg ($0.50) and leftovers. Yes, there's packaging and stuff as well, but for
that price they could equally well have made it a kilogram.
On the way home, dropped in at Bunnings in Delacombe, apparently the biggest
shop they have. Once again, it was almost empty, and the sole cashier explained this:
That's supposed to be the do-it-yourself checkout, but it's closed: they need a person to
supervise it, and it's just not worth the trouble with the amount of business they have. I
wonder how things will pan out. Presumably people will learn about the place in the course
of time.
The shop is laid out in a very similar way to the
other Ballarat shop. Off to look at
floor tiles for the verandah. Didn't find very much, but did find this:
OK, $45.77, and the tiles are only 60 cm square. That's 0.36 m². But the small print
states $42.36 per SQM. How does that work out? Must be bad language, like TILE
FLOOR PRCLN PEAK CTN3. And how they get 1.08 SQM out of a tile 60 cm on a side
is hard to understand. After a while, it dawned: CTN3 means “carton of three tiles”,
and the price is for the carton, not the tile.
Wouldn't it be easier if they wrote these things in English?
And then there are “smart” plug-in sockets, a pair for about $27:
Arguably they're useful; at least they don't rely on a smart home controller. But
they do rely on a smart phone. Why? Clearly they communicate
by 802.11, so anything on the network
should be able to communicate with it. Sigh.
Finding my way out of the Bunnings car
park wasn't easy. No signs apart from a couple of no-entry signs where I didn't need them,
various barriers. In the end I decided that the only exit was to the south onto Webb Road,
the way I went last time.
OK, I wanted to go south on Cherry Flat Road. But other things prevented that:
They were blinding to the eye, almost completely obscuring the “Road Closed” message that
they tried to convey.
OK, Google, how do I get home? Oh, that's an old, worn-out phrase. Now it's “Hey Google”.
And as usual the interminable delay before this supercomputer in my pocket finally does
something, showing me the way home (straight ahead at this point). Drove on, took another
look, and got what looked like a scrambled screen:
Had I moved? I don't recall. The GPS log says yes, by 5 metres, but the map claims 100 m.
Still, what kind of display is that? I can't see anything there. More messing around and
gone nothing useful, just spam from Google:
I suppose I should check out the app, but I suspect it will just give me more cause for
grumbling. Currently I download the screen shots to a real computer and edit them, if
necessary, there. Anything that requires me to spend more time on a mobile phone has to be
bad.
Huh? And all this, of course, with no verbal feedback from the device. Dammit, Google,
take me home!
Still more messing around and I discovered that, for some reason, it had tried to send me by
bus. Probably I prodded the phone in the wrong place; that's far too easy. Got to a T
junction in Sebastopol,
which I recognized, and somehow managed to get it to show me the map I had been looking for
all along:
OK, Google? It's Hey Google! And I certainly don't want you to play me any jazz. But
after that it navigated relatively well with the exception of when it asked me to turn left
into an embankment going
through Enfield State Park.
That reminds me of last
year, where it told me to turn right when going along a road with no turnoffs. I
think Google Maps must not like forests.
Yvonne in this afternoon to point me at a flower growing in
the “sewage paddock” (the one with the soakage for the septic tank). She thought it might
be an orchid:
No, it's clearly not an orchid. But what is it? It reminds me of something, not
necessarily indigenous. OK, time to look for programs that can identify flowers.
I didn't find any! Only “apps” for mobile phones! What a declaration of bankruptcy! What
happened to our ideals
of interoperability? And what a
horrible thought to have to upload my photos to a phone to be able to identify them?
Sure, it's convenient to be able to take a photo with a phone and identify it immediately,
but a good program shouldn't be platform-dependent.
Later, while walking the dogs, found another flower:
Our Leucadendron salignum
cultivar, the one that Lorraine Carranza gave us last Christmas, has been here for nearly a
year, and so far it hasn't flowered. But finally it's promising:
Looking around the verandah area today, discussing the loss of the goldfish. And then one
showed up again, like the one I saw a couple of weeks ago. And another. And another. All
in all I counted at least 5 of them, all relatively big.
What's going on? Are they just hiding? Or are they somehow sick or undernourished? Or
maybe hiding from a predator that really tried to catch them? Tried feeding them with the
flakes that Yvonne bought for them. One of the biggest ate
some, but the others didn't pay much attention. In any case, time to clean out the trough
and eliminate some of the plants that have grown in the water.
I've never made samosas before, and today I decided to try it out. I discovered multiple
issues:
Samosas are deep-fried, not baked.
The instructions want you to heat the cooker (to a surprisingly accurate and uniform
180°, it eventuated) and then roll out the pastry onto the hot plate, fill them in
situ and cover with another layer of pastry. I don't know how well that would work,
but at the very least it would be wasteful of pastry. The results would look nothing
like the illustration.
The instructions also suggest oiling the surface of the cooker to prevent sticking.
That sounds like it could be a lot of fun removing the samosas.
OK, first learn to make samosas the traditional way. We have forms that make pouches in
samosa size, even if they're semi-circular. They don't waste pastry, and they don't sound
so time-critical. I'll try that first, and if they taste good, I can go looking for
something that makes triangular rather than semi-circular pastries.
Waling the dogs today, saw another kind of wildflower:
It's not easy to recognize, just small (1.2 cm?) white flowers growing apparently from
nothing. The interesting thing is that they don't have a constant number of petals:
For a few months now I have been converting my intermediate photos
to TIFF rather
than JPEG. And of course it's bloating the
backups. Today I removed most of the intermediates. The difference is obvious. Here
before and after:
Yvonne took a few photos this afternoon, about 430 of them.
She zooms with the crop tool, so she was using dischord all afternoon, and I had to
use eucla, the laptop. It doesn't have the latest and greatest Ashampoo optimizer, so I tried the old 2016 version that
had broken on me last
month. It didn't fail!
But when comparing the results, though they were marginally bigger, there was no obvious
difference in appearance. Optimizer fail. Later I ran the same photos through “Photo
Optimizer 7”, and it did improve them. What's going on here?
Another email from Resilium again today.
No, this time it was “Non Fleet Recoveries” (and not ”Fleet Recoveries” like the last time).
It was signed by “Arpana”, of whom I am not sure whether she is the same person as Amol or
not.
What did they want? $550. The offer is still open, but if I don't pay by 3 October they'll
forward it to their mercantile agent.
What's wrong with this picture? No mention whatsoever of my reply of 27 September. Are they unable to read? Certainly they have given the impression of having far less
intelligence than even typical insurance representatives, and the date 3 October adds to
that impression. Maybe I should just pay the $550 to be rid of the pain.
Deferred the “CONTACT ROASTING MACHINE” and looked at the laptop, which looks quite good.
Problem: there's something wrong with the display, and from time to time it becomes more or
less illegible. Repair will apparently exceed the value of the machine. For me that's not
a significant disadvantage, since I'd want to access it over the net only.
Took it into the office and opened it. Well, tried to open it. Dammit, I can't get it
open! Is this another silly Apple trick? Off to the web, where I found precious little
help beyond the suggestion that some of them can be a little stiff. Insert big screwdriver.
Nothing.
After a while, looked at it more carefully and discovered that I had been trying to open the
wrong side. The other side opened quite easily. If there was a silly Apple trick, it was
in putting the logo on upside down and putting all the connections on the sides, so that it
wasn't clear which way was front and back. Andrew Perry told me that all logos are upside
down nowadays, such as this ThinkPad
T460, from this page:
But I have a ThinkPad too. The one above is a T460, and mine is a T430:
Clearly people aren't as clear about orientation as they claim. And clearly having
connections on the back makes it more difficult to make the mistake that I made.
OK, fire up. How about that! I had an account on the box, from who knows when? Logged in
and was presented with the offer to do a system upgrade. OK, it proved the machine hadn't
been used for nearly two years, so why not?
Aaargh! This bloody Apple ID! It's set to Chris' ID, of course. I had an ID, and maybe I
still do, though I went through untold pain to get rid of the invasive thing earlier
this year. Should I reinstate it?
First, how? The popup says “If you don't have an Apple ID, click Create Apple ID”. But
where? It seems that it forgot to add that button!.
OK, Cancel. Ha ha, only joking, you can still download the new system. Only 5.3 GB.
Left that running for some time, and addressed the issue of running the laptop when it's
closed: normally that causes it to hibernate. In Microsoft there are buttons to set that,
but Apple is too polite. Nothing in “Energy Saver”, nothing that I could find elsewhere.
SCREAM!PAY for programs that make up for brokenness in the base
system! What a horrible idea!
More looking and came up with Quick Tip: How to Stop Your Mac From Sleeping Using the Command Line.
Use caffeinate(1), which does come with the base system. But what an appalling
kludge! It stops the machine from sleeping by running a make process (and doesn't
supply a Makefile, making it even more difficult. I'm appalled!
At 10 Mb/s, 5.3 GB take about an hour. In the meantime, tried something different. For no
good reason, compared the ping times. At this point the system had called
itself dhcp-238. Here a summary:
--- teevee.lemis.com ping statistics ---
round-trip min/avg/max/stddev = 0.142/0.311/0.374/0.069 ms
--- lagoon.lemis.com ping statistics ---
round-trip min/avg/max/stddev = 0.315/0.364/0.402/0.027 ms
--- dhcp-238.lemis.com ping statistics ---
round-trip min/avg/max/stddev = 0.632/0.788/0.926/0.012 ms
Why so slow? It's a 1 Gb/s interface, with only one switch between it and eureka,
while the others had two.
Finally the upgrade was done. Or was it? It came up with a conflict
with Xcode, without a suggestion as to how
to fix it.
OK, time for another screen shot. But why with a camera? Apple must have some way to take
a screen shot. More checking. Went out to the web, where I discovered that the way to take
a screen shot depends on the version of Mac OS. What version of Mac OS do I have currently
running?
dhcp-238:~ grog$ uname -a Darwin dhcp-238.lemis.com 13.4.0 Darwin Kernel Version 13.4.0: Mon Jan 11 18:17:34 PST 2016; root:xnu-2422.115.15~1/RELEASE_X86_64 x86_64
That doesn't help much. How old is the kernel?
-rwxr-xr-x@ 1 root wheel 8394688 Jan 12 2016 mach_kernel
That doesn't look like a recent update. Was there something about a Sierra? Is that the
same as High Sierra? How does that relate to other OS revisions? How do I find out?
Apple support notes, of course. On IRC, Jamie Fraser offered me this page, which goes into some detail.
But you need macOS Mojave (is
that the same thing as Mac OS?). They're too polite to tell you where to find information
for other, presumably older and mouldier versions of their name-changing operating system.
Finally got something that suggested that Cmd-3 will take a shot of the entire
screen. Entered Cmd-F3 and ended up with a completely different screen:
The word “purchase” sounded ominous. Is this a polite way of circumlocuting “You need an
Apple ID and $$$ to buy this upgrade”? OK, they offer Details:
OS X Mavericks? Is that the same as Mojave? Looking at the dates and the comment “This
article has been archived and is no longer update by Apple”, I'd guess that this relates to
a much older version, and that the link that took me there is broken. Bad Apple! Still, I
don't see myself spending real money buying “Dark Mode” (presumably the horrible colours
that my photo software offers, and which I've been grumbling about for I don't know how
long), “new apps”, and most certainly not a new Mac App Store. Stacks? What's that? Do I
want to know?
In the meantime I got a popup telling me to update Java:
But it seemed to hang. OK, go to “About this Mac” and generate a report. Screen scrambles,
the problem that made Chris stop using it. What she didn't know, though, was that it wasn't
just the screen: the system froze and was no longer accessible via the network.
That's enough pain for one day. Why did I do it in the first place? Clearly Apple and I
are so far from each other that nothing can save the relationship.
Yesterday's fun with the MacBook should
have been enough, but there were still some loose ends to tie up. What is this machine?
Were any of the 5.3 GB of updates yesterday actually installed? After reboot, I found these
displays:
What does the X in the circle mean? That the update failed? Or that this is Mac OS X? It
seems that that “Mac OS X” is an old, worn-out magic word, and that it's now
called macOS. But I've already seen that
they're not very consistent. After all, you'd really think that they're practicing
deliberate obfuscation by having no less than three names for their releases. If I can
believe what it's telling me, I have Darwin 13.4.0, OS X 10.9.5 and macOS High Sierra. Do
they relate? My kernel (if that's what /mach_kernel is) dates from January 2016, and
I can't see any updates to /usr/bin (nor even find/sw/bin, which
appears in my PATH variable, whatever that might be intended to mean).
Other details are scanty. 16 GB memory (good, more than any other machine I have except
for eureka). What processor? Ah, we won't scare you with details like model number.
It's a 2.5 MHz Intel Core i7. How many different kinds are there of that? According to the link, there are at least
12 of them, assuming you discount the 2.53 GHz processors (yes, that's a different value,
but maybe Apple is too polite to distract me with more than 2 significant digits). How many
cores? There could be 2 or 4.
And the OS release? This Apple
support note tells me how to find out the macOS version number:
From the Apple menu in the corner of your screen, choose About This Mac. You'll see the
macOS name, such as macOS Mojave, followed by its version number. If some product or
feature requires you to know the build number as well, click the version number to see it.
It was badly placed, thus the poor shot, but hopefully I'll see one again.
And then this shrub is directly in front of the house. Yvonne transplanted it from the other end of Stones Road. It's the only one of three that
survived, and it looks better than any of the originals, which are normally even stragglier.
Chris Bahlo has finally got round to updating her data on the new web server
machine—with SFTP! I
wonder how she keeps track of things. And of course it didn't work for her, because I was
still running the zone on the old web server.
OK, no time like the present, even if it is just before dinner. Copy across the information
and run apachectl graceful. Error! The pathnames were a kludge that I needed on the
old server, /usr/home/chris/www.narrawin.com. We don't need to put /home
under /usr any more, so /home/chris/www.narrawin.com it is.
OK, apachectl accepted that. And then I had to wait for the DNS to propagate.
Finally it happened, but I couldn't access the site. The error log contained:
[Tue Oct 08 07:24:54.084811 2019] [authz_core:error] [pid 89695] [client 167.179.139.35:61970] AH01630: client denied by server configuration: /home/chris/www.narrawin.com/h3
[Tue Oct 08 07:24:54.433312 2019] [authz_core:error] [pid 89695] [client 167.179.139.35:61970] AH01630: client denied by server configuration: /home/chris/www.narrawin.com/favicon.ico
What does that mean? apachectl gave it a clean bill of health. Took a look at the
configuration again
We've had our Grevillea
bronwenae for two years now. It looked OK in the first year, but in the second I was
marginally concerned. But this year it has come back with a vengeance. Here two years ago,
last year and today:
So was Nikolai's problem last
week sickness or sadness? His bowel movements have long since normalized, so today I
tried it again. Go walking without a chain or leash. No, he didn't want to know, wanted to
go back inside, just like last week. Put him on the leash and he was happy.
My pains with Resilium or FleetRecoveries
or NonFleetRecoveries aren't over yet. Yesterday I sent them a message in the assumption that their
attention span doesn't extend beyond the first paragraph, and asking at the end of it for a
confirmation of receipt. None was forthcoming.
OK, from now on I'll send them a message every day until I get a response. I wonder if I
ever will. What an incompetent company!
I've already established that the ALDIsamosa maker appears to be suboptimal.
But what's the best way to make them? Not surprisingly, I found a number of recipes both in
my recipe books (one for “samoosa”) and on the web. All different. It seems that you can
make them with vegetables or with meat, but vegetables seem more appropriate. Today spent a
lot of time trying to find a recipe, and came up with this:
quantity
ingredient
step
5 g
ginger
1
36 g
onion
1
0.5 g
black mustard seed
2
2 g
coriander seed
2
1.5 g
cumminseed
2
1 g
fennel seed
2
2.25 g
garam masala
2
130 g
cooked potatoes
3
60 g
peas
3
1 g
curry leaves
4
5 g
coriander leaf
4
2.5 g
salt
4
Preparation
Cut the onions and ginger finely, fry in a little oil until onions are glassy.
Grind the spices and add to the onion mixture.
Add potatoes and peas and crush with a potato masher.
Add leaves and salt and mix well.
Roll out shortcrust pastry and cut into circles. Fold together in a wonton press.
Deep fry quickly at 180°, then complete at a later time at round 150°.
That was very much an experiment, but it worked surprisingly well:
I'll need to frob the quantities next time,
but the interesting thing is that the relatively small amount of filling was enough for
about double the quantity in the photo.
Another grid power failure this evening at 18:44:25, again only one second. I wonder if this kind of failure would
have registered before we installed solar electricity.
That was repeatable. What now? The host name was
interesting: verteiler1.mediathekview.de. Is there a verteiler2? Yes.
And it works!
Further investigation shows that there are a number of them, not all distinct:
=== grog@eureka (/dev/pts/33) ~ 57 -> for i in `jot 10`; do host verteiler$i.mediathekview.de; done verteiler1.mediathekview.de has address 5.1.76.111
verteiler1.mediathekview.de has IPv6 address 2a00:f820:417::4df6:1bf2
verteiler2.mediathekview.de has address 195.201.115.158
verteiler2.mediathekview.de has IPv6 address 2a01:4f8:1c0c:7baa::1
verteiler3.mediathekview.de has address 5.1.76.111
verteiler3.mediathekview.de has IPv6 address 2a00:f820:417::4df6:1bf2
verteiler4.mediathekview.de has address 88.99.80.17
verteiler4.mediathekview.de has IPv6 address 2a01:4f8:1c17:5753::1
verteiler5.mediathekview.de has address 5.1.76.111
verteiler5.mediathekview.de has IPv6 address 2a00:f820:417::4df6:1bf2
verteiler6.mediathekview.de has address 159.69.82.180
verteiler6.mediathekview.de has IPv6 address 2a01:4f8:1c1c:5642::1
Host verteiler7.mediathekview.de not found: 3(NXDOMAIN)
Host verteiler8.mediathekview.de not found: 3(NXDOMAIN)
Host verteiler9.mediathekview.de not found: 3(NXDOMAIN)
Host verteiler10.mediathekview.de not found: 3(NXDOMAIN)
=
Walking the dogs round the “big loop” (Rozenstein Road), came across a number of
wildflowers, including some that I don't think I've seen before. Now is clearly the time
for the Caladenia major:
One of Yvonne's big gripes about Real Cameras is the lens
cap. I can't blame her. Somehow, after round 100 years, the industry still can't agree on
how to make one. Today I took the M.Zuiko Digital ED 60 mm
f/2.8 Macro with this lens cap, which I think originally came with the Leica Summilux 25 mm f/1.4:
And the bloody think kept falling off! Why can't they make a cap that's easy to take off
and put on, and that just stays on? 50 years ago the lens caps were just push-on, and I
can't recall ever having lost one.
Piccola into the house today with a treasure
found in the garage. Yvonne was surprisingly upset, and I
locked Piccola in the laundry. But she wasn't overly interested:
Yvonne has borrowed a number of USB sticks from Julie Lannen
which she wanted copied. All Microsoft format, of course, one of which eureka didn't
want to know about. It proved to be in NTFS
format, not a good idea for interchangeability, and I had to read it in on dischord.
What a pain this Microsoft is! And all these file names with spaces in them; three even had
leading spaces.
What really interested me, though, is the attention people pay to these things. Each stick
was different, and two of them were in a decorated case:
Yvonne had an appointment for some kind of scan today, in
principle nothing that would involve me. But yesterday she had had a couple of incidents
where the starter of her car didn't engage, clearly something that we needed looked at. So
off with her into town and left the car at what used to be Ballarat Automotive, but which is
now under new ownership and called Sovereign City Service Centre.
To my surprise, Leigh (Franklin) was prepared to look at it right away, and called me to say
yes, needs a new starter motor—probably not surprising for a 2007 model. And to my
surprise, the motor only cost $250, and he could have it ready this afternoon.
That wouldn't work, since we'd be home by then, but Yvonne is going into town again
tomorrow, and Leigh is more than happy to be at the workshop then to return it to her.
After dropping Yvonne at St John of God I went on to
the East Asian food shops in Howitt St. Found some onion shoots, something I've never seen
before:
Then to the Pilipinos, where I almost never find anything. I certainly didn't find
everything I was looking for, but they did have Californian Sambal “Oelek” and some Bolsts
Indian chutneys. In particular I hadn't expected the latter.
After that dropped in to the Botanical Gardens, where nothing much seems to be happening. The Friends of the Ballarat Botanical Gardens are now asking
$8 for the plants on sale outside the Robert Clark Centre, a price that I can resist. Took
a look inside the centre, which seems less active than in the past. I don't know if it's a
good thing or a bad thing that there are no people in there any more:
After that on the the “new” (for me) barber that I had been intending to try out, the
King's Crown. Surprisingly
active place, old-fashioned appearance, 3 barbers cutting in semi-darkness. And a 25 minute
wait! Looking at reviews on the web, this seems to be the short end of the wait. I might
have tried it, but there was every chance that Yvonne would
be finished before I even got into a barber's chair. Looking at their web site, I discover
that they also cost twice as much as Kerry
in Sebastopol.
So across the road (almost) to Formosa Gardens nursery, where I found a couple
of Fuchsias for the hanging pots that we
must finally hang in the entrance. Also bought a chili bush and a Thai basil bush, which I
hope will grow better than the European basil we've tried in the past.
Eucalyptus
dolichorhyncha, misspelt as Eucalyptus dolichorrhyncha, and not quite the Fuchsias that
I had found, but possibly interesting. To be investigated.
I've recently grumbled about problems updating Apple software. In many ways it seems that
Microsoft does it better.
But Microsoft is up to the challenge. I started an update yesterday—only a few hundred
megabytes, something that I could download in 5 minutes—and today it was still hanging at
48% downloaded. Restart. From 48%? No, from the beginning, of course. I sometimes wonder
what the issues are with their downloading process. Anyway, it worked this time.
It's been four
days since I have asked Resilium or
FleetRecoveries or NonFleetRecoveries to simply confirm that they had received my email.
Did I get one today? No, not explicitly. But I did get another that might almost make up
for it:
Date: Fri, 11 Oct 2019 03:46:51 +0000
From: Fleet Recoveries <FleetRecoveries@suncorp.com.au>
To: 'Greg 'groggy' Lehey' <groggyhimself@lemis.com>
Subject: RE: Letter of Demand | Claim Number: R005112442
Message-ID: <BA8C59D57C1F65429ACF4AA8F536774D527ECC58@PBNEMBMSX4107.int.Corp.sun>
We are willing to accept the payment of $300 to settle the matter.
I have attached the previously sent letter for demand for payment options.
That should really make up for things. I'm left wondering whether I should have offered
less: clearly they don't have any evidence against me. But I had expected it to cost $300,
and any less could have ended me up in litigation. I would almost certainly have won, but
as I have already noted, it's worth $300 to have nothing more to do with these idiots.
I've established that one of the reasons that I have so much web traffic is because I have
multiple names that respond to HTTP
requests, so crawlers load each page several times. Apart from www.lemis.com,
there's also mail.lemis.com, people.lemis.com and lemis.com. How do I
tell the client to redirect to www.lemis.com?
Out searching the web, and came up with myriad way to do similar things, including
with mod_rewrite, like this page from the the
Apache documentation. But
further down it came up with the answer: a simple Redirect directive. Just
what I was looking for, but why did it take me so long to find it?
So put that in the configuration, which also greatly simplified it:
Load average 80! I've never seen that before on the old server. In fact, there's an
indication that it passed 100:
last pid: 79336; load averages: 19.30, 16.94, 18.50821 up 47+16:28:00 22:12:34
Those 5 digits after the last average suggest that the text had once been stretched that far
due to the length of the number.
Was that because of the redirects? Shut down the server on oldwww (the old machine)
and watched the load average drop to under 3. Why so many?
The other thing was the 302
code. What does it mean?
302 Found (Previously "Moved temporarily")
Tells the client to look at (browse to) another URL.
In other words, go elsewhere this time only. Not what I wanted. More searching and
found this page, which told me what the Apache documentation didn't: Redirect
takes an optional first argument (shifting the others right). What I wanted was:
Redirect permanent "/" "http://www.lemis.com/"
Put that in after I didn't need responsiveness from the system. Hopefully things will calm
down by tomorrow.
After doing yesterday's photos, ran my normal file sync script. Things didn't go exactly
according to plan:
rsync: mkdir "/home/grog/www.lemis.com/grog/Photos/20191011" failed: No space left on device (28)
rsync error: error in file IO (code 11) at main.c(670) [Receiver=3.1.3]
rsync /home/grog/public_html/photos/dirlist www:/home/grog/www.lemis.com/grog/photos
rsync: write failed on "/home/grog/www.lemis.com/grog/photos/dirlist": No space left on device (28)
rsync error: error in file IO (code 11) at receiver.c(381) [receiver=3.1.3]
Sat 12 Oct 2019 13:50:00 AEDT
It took me a while to realize that the problem was at the remote end. That box has so much
more storage than I need! Took a look and finally found my HTTP server error log.
42 GB! I should be rotating the logs, but there are issues with Apache log rotation
that I haven't investigated yet. Still, what's in there? The ones at the end were
particularly helpful:
[Sat Oct 12 01:47:46.802988 2019] [log_config:warn] [pid 42148] (28)No space left on device: [client 64.62.252.176:42598] AH00646: Error writing to /var/log/www/lemis.com.log
[Sat Oct 12 01:47:46.839821 2019] [log_config:warn] [pid 42048] (28)No space left on device: [client 1.31.97.14:37774] AH00646: Error writing to /var/log/www/lemis.com.log
[Sat Oct 12 01:47:47.235456 2019] [log_config:warn] [pid 42058] (28)No space left on device: [client 123.134.255.141:56777] AH00646: Error writing to /var/log/www/lemis.com.log
Apart from that, lots of warning messages
that PHP was previously too polite to report,
some from scripts that should no longer be used. Lots of people trying to access Yvonne's photos, which used to be on the server, but are now on
DigitalOcean. OK, remove the empty
directories (things like /yvonne/Photos/20180827/big/, and think later about how to
bend the 404 document to point to the correct place. Remove the old script and link to its
replacement. And start fixing the more obvious problems with my scripts. A lot were the
unquoted use of $ in text input to a PHP function. And after a while the error log
started growing less quickly than the access log. The rest can be done later (how many
times have I said that?).
Chris Bahlo for dinner this evening. She was a bit stringy; we should have marinated
her. Once nothing worth mentioning: she used to come every Saturday. But times have
changed. Spent much time talking about the strange situations many of our acquaintances
have got themselves in to.
Amazing information about my grid power today. Between 0:31:26 and 0:33:48 we had no less
than 11 grid power failures.
Clearly they weren't long: the total period was 2 minutes, 22 seconds, and the longest
outage was 27 seconds. On the other hand, this kind of bouncing is hard on electronics,
probably the reason that the inverter waits a minute before reconnecting the grid.
But how do I count this? One 3 minute outage would be longer than the 11 I had. For the
time being I'll count it as individual outages.
Could it be that these were brownouts? There have been significant voltage fluctuations,
but not at that time:
Of course, that's only part of the truth. Clearly the grid voltage dropped below an
acceptable value during that time, but there's nothing whatsoever to show it in the graph.
What really happened?
It has been almost exactly 10
years since we visited
the Brisbane Ranges, specifically
to visit the Friends of Brisbane
Ranges Wildflower Show. Unfortunately they seem to have stopped doing that, so this
year we decided to go it on our own, specifically to the east end of Butchers Road, which
runs between Steiglitz
(pronounced Stieglitz)
and Anakie. We had found the
most flowers there 10 years ago, and we weren't disappointed today. Directly next to the
car we found two difference species of orchids (at the extreme left and right):
But which Diruis? Diuris pardina?
There are some similarities (of course), but also potential differences. Are these two even
the same species? My guess is no.
For the time being I've given up trying to name most of the flowers; I'll try to identify
them and come back here. Here are some that I consider to be orchids:
On the way back, passed tracks marked “Wildflower Trak” and “Orchid Trak” on my GPS
navigator, which seems to have a predilection for the spelling “Trak”. Stopped at the
Wildflower Track, though Yvonne had had enough, and walked down a bit. Yes, wildflowers,
but nothing that I hadn't seen at the other place. Met some women who told me that they had
seen a spider orchid, and where it was, but all I saw there was one of the Grevilleas I had
seen.
The way home went
through Meredith, a
not-so-small town on the road
from Ballarat
to Geelong. I've been through several
times, but never stopped. The weather was good, so today seemed like the day to try it,
especially as we discovered that you can buy ammunition and old drinks at the Post Office:
After the extremely high load averages of the last few days, www.lemis.com is now
returning to what I hope will become normal, with typical load averages below 1, though from
time to time it goes up to 20.
But barely has that problem gone away that I get another:
From: "support@vultr.com" <support@vultr.com>
To: groggyhimself@lemis.com
Subject: Vultr.com: Bandwidth Warning
The following subscriptions have reached high bandwidth utilization levels:
Subscription:3243454 - 2048 MB Server - 66.42.97.229 (w3.lemis.com): 78% used
...
Please note: Your bandwidth usage cap will reset on the 1st of every month.
Oh. 78% only 13 days into the month? A result of the recent load issues? No, it seems
that the bandwidth load has been relatively even. More head-scratching to do.
I don't know how long it has been since we decided to hang some baskets on the left in the
house entrance, but it must have been round a year ago. Finally, with the advent of warmer
weather (again!), got round to planting
the Fuchsias that I bought last week:
Surely that must be an orchid. But what kind? There are even more of them here than there
were Caladenia major
in Steiglitzyesterday.
And once again I had difficulty keeping the things still. There must be some clever way to
keep flowers from shifting in the wind without damaging them or making the support evident
in the photo.
So, as everybody (except, possible, Donald Trump) had expected, the Turks have invaded
northern Syria and attacked
the Kurds, allies of
the USA. And where is the USA? Its supreme
leader is pretending it has nothing to do with him:
Sadly, this image has disappeared, and I had forgotten to make a copy. The URL
was https://pbs.twimg.com/media/EGf7Uk4XUAAzK1x.jpg, which doesn't give much to
search on.
With friends like this, who needs enemies? What a message to send to the rest of the world!
This will really make negotiations with Iran
and North Korea so much easier! As
one commentator put it:
Faced with a crisis of its own making, a flailing superpower has turned to economic
sanctions to pretend it is still relevant,
That wasn't just anybody. The Economist is
one of the most influential news sources in the Western world. What they're basically
saying is “Trump has destroyed the USA”. Make America Great Again? Hah! The article goes
on to suggest that the real solution lies in the hands
of Vladimir Putin.
Four years ago I considered, and ultimately bought, a wide aperture standard lens for
my Olympus OM-D E-M1. My choice was the Leica Summilux 25 mm f/1.4. It wasn't the widest aperture I could get, but all the
ones with wider apertures didn't have autofocus. Since then the M.Zuiko 25 mm
f/1.2 has appeared, and it has autofocus, but even at the time there were lenses with
apertures as large as f/0.85—only all with manual focus.
I bought the Summilux before the Olympus lens came out, but I don't think I would have
changed my decision. I don't use the lens that often, and the price and size differential
is significant. On the other hand, a 25 mm f/0.95 lens with autofocus would have been
interesting.
Now Nikon has come out with its own standard lens with f/0.95 aperture, for the Z series.
For some obscure reason it's 58 mm instead of the more normal 50 mm. It's enormous and
prohibitively expensive. Here a comparison from Compact Camera Meter:
That's a Nikon Z7 “full frame” camera, barely larger than the E-M1 Mark II. But look at the
monster lens! “Full frame” bodies can really now be the same size as Micro Four Thirds system bodies, but the lenses are generally larger.
The real issue, though: the new Nikon lens is manual focus! Why? It can't be the cost of
the thing: they're talking about US $8000, more than I paid for all my bodies and standard
lenses put together. By contrast
the Voigtländer 25 mm f/0.95 lens for
MFT costs $800, exactly 10% of the cost of the Nikon lens. Is there some technical reason
for the lack of autofocus? In general autofocus works better with wider apertures, and
that's also where you really need good autofocus. I'm puzzled.
Ashampoo is currently celebrating their 20th
anniversary. And I'm still having trouble with them. Their latest optimizer, version 7,
works, but it seems to impart a greenish cast on many images. And version 6 no longer
refuses to run: it now creates output files that are round 40% larger than the original, but
otherwise appear unchanged. If I select a single image, it still fails.
But that's nothing new. By chance I looked at my diary of 5 years ago. Same problems then. Why do I stick with them? Why doesn't somebody else come out with
something similar but more reliable?
Yvonne planted
some Artemisia cuttings at the
north-eastern corner of the property a while back. They all died.
They're not the first thing to die there; in fact, everything that we planted there
has died. But Yvonne suspected that they weren't getting enough water. OK, out to check.
First problem: somehow Mick had managed to disconnect a hose connection on the eastern
boundary, resulting in no water for a lot of plants to the north, including
the Paulownia kawakamii, the
remaining birch and the oak. But they're all doing well. Here a flower from the Paulownia,
the first we have had since moving here:
OK, fix the join and back to the row of Artemisia. Opened the end to let the water out, a
good idea in any case: sludge forms in the hoses in the winter, and it's better to flush it
out before it clogs the drippers. But nothing came out.
Further investigation showed that the hose is on relay 3, not relay 4. That works fine, and
it has been doing so the whole time. So the whole premise, while it made me check the
irrigation and find faults, had nothing to do with Yvonne's concerns. We'll have to start
again with new cuttings.
It's been pretty busy the last few days, to the point where I haven't got round to doing a
number of routine things. Even this diary hasn't been ready until mid to late afternoon.
My RCS log times show the
effect:
They're in this silly reverse chronological form, and in UTC; my time is 11 hours
later. So instead of a typical 10:30 (bottom line), the times have gone up to 16:23.
Today I relaxed, and found time to plant the cuttings from the mystery plant that I took
yesterday:
From time to time we've eaten some
instant Miso soup, not the most
fascinating soup I know, but it's a usable addition to
a Sushi/Sashimi
dinner. It's boring enough that we have nicknamed it „Miese Suppe“ (roughly “terrible
soup”), because that sounds pretty much the same. But why instant? One step better would
be pre-prepared broth in a bag, from Wokka. As the bag said, just add
water, Ramen noodles and spring onions.
So we did that. It tasted terrible! Well, even more boring than the dried stuff.
One issue was the quantities: the instructions specified 300 ml of water, and that was
clearly too much. We ended up adding soya sauce to get any kind of taste. We could equally
well have used water and forgotten the concentrate.
Another 5 short grid power
failures this morning between 07:45:27 and 07:46:26. That's less than a minute.
Should I really count them individually? In particular, it seems to take the inverter 1
second to register a power outage, so if there's only 1 second between the failures, it's
not really power back at all. Maybe I should set a minimum of 5 seconds between outages.
Another all-day “scheduled” outage from the National Broadband Network today, from 8:38 to 16:29,
nearly 8 hours and almost exactly the whole working day. Spent some time hacking my
network failures program to produce more
interesting information, like:
Summary from 1 January 2019 to 31 October 2019
Total 64 outages, total time 195460 seconds (2 days, 06:17:40)
Longest outage: 28252 seconds (07:50:52)
Start: 17 October 2019 08:38:34
End: 17 October 2019 16:29:26
Average time between outages: 392149 seconds (4 days, 12:55:49)
Average duration: 3054 seconds (00:50:54)
Availability: 99.22%
So it's the longest outage this (very bad) year, and the tendency seems to be getting worse:
Summary from 1 July 2019 to 31 October 2019
Total 31 outages, total time 139734 seconds (1 days, 14:48:54)
Longest outage: 28252 seconds (07:50:52)
Start: 17 October 2019 08:38:34
End: 17 October 2019 16:29:26
Average time between outages: 305017 seconds (3 days, 12:43:37)
Average duration: 4507 seconds (01:15:07)
Availability: 98.52%
Average outage duration over the last 3½ months of an hour and a quarter,
availability of only 98.5%. I've been using the Internet for over 30 years now, and I can't
recall ever having had such unreliability. Even my satellite link had an availability of over 99%, and the outages were shorter:
Total 1684 outages, total time 792001 seconds (9 days, 04:00:01)
Average time between outages: 56449 seconds (15:40:49)
Average duration: 470 seconds (00:07:50)
Availability: 99.17%
It's made all the worse by the fact that it's deliberate.
What are they doing? Preparing us for 100% down time, so that they can delight Wendy
McClelland's heart and remove the Radiation
Tower? It seems so. This particular outage was part of a wider announcement:
Date: Fri, 27 Sep 2019 12:20:07 +1000
From: Aussie Broadband <support@aussiebroadband.com.au>
Subject: NBN is planning some maintenance in your area
NBNCo has let us know that they are planning network maintenance in your area, and that your service at <strong>29 STONES RD, DEREEL VIC</strong> will be affected.<br>
The details are:
- Start date and time: Wed 16th October 2019 07:00 AEDT
- End date and time: Tue 22nd October 2019 21:00 AEDT
- Window: 158.0 hours
You may experience the following interruptions during the maintenance
- 720 min
- 720 min
- 480 min
- 480 min
That's a total of 40 hours, a whole working week! Our outage today was presumably one of
the first two. Do they really plan to take us off another three times for hours on end
between tomorrow and Tuesday?
After the day's outage, recovery took a while. For some reason my outgoing mail tunnel
to mail.lemis.com didn't want to restart automatically. When I started it, it hung.
Oh.
#!/usr/local/bin/bash
# Set up and maintain mail tunnel to mail
# $Id: diary-oct2019.php,v 1.50 2024/10/22 03:23:21 grog Exp $
while :; do
logger Restarting SMTP tunnel
/usr/bin/ssh -n -N -L 2026:mail.lemis.com:25 www.lemis.com
sleep 5 # don't flood
done
For reasons I forget, mail.lemis.com is 192.109.197.81, one of the last uses
of my /24 network block, and an alias for the interface on the old www. But I had
had difficulties with routing, which didn't always work, so I had made the actual connection
to www. And that has changed, and mail on www is not (yet) configured.
Changed to oldwww and all was well.
And then I had difficulties connecting to www with ssh. A lot of flailing
around, and finally I got a connection. But that load average!
That's worse than it ever was on the old server. What's the issue? I had thought that the
load was due at least in part to the multiple names under which the server was known. Now
most of them have taken the 301 hint, but some remain (semrush.com, are you listening? Probably not). Is the
new server slower than the old one? Certainly it has only one logical CPU, where the old
one has two.
Looking at that and also the amount of traffic I'm getting, it looks as if it might be
cheaper to migrate to a more powerful vulture. Currently I have a single processor with
2048 MB memory and 2 GB bandwidth for $10 per month—plus possibly $20 in additional traffic
charges if I really use the 4 TB that it's looking like. But for $20 total I can get a 2
CPU vulture with 4 GB memory and 3 TB traffic. Might that not prove cheaper in the long
run?
The effects of the overload were evident when I tried to sync my diary. Normally it takes
about 30 seconds, but today I got:
sent 47,684,649 bytes received 198,394 bytes 14,529.83 bytes/sec
total size is 3,109,981,811 speedup is 64.95
Thu 17 Oct 2019 18:13:44 AEDT
real 55m4.814s
user 0m9.193s
sys 0m0.253s
Why did I transfer nearly 50 MB? Probably some temporary log files that I created when
updating the failure program. But it appears that the transfer completed relatively
quickly, and it then hung. I don't have an explanation for that particular behaviour.
The tiny flowers on our new mystery bush are about 4 mm in diameter and 1.5 cm long. But
they have an internal structure, if only I can get it. Tried today with the M.Zuiko Digital ED 60 mm
f/2.8 Macro and extension tubes, for an on-sensor magnification of 1.4. That required
focus stacking, of course. This one didn't make it either, although I took 60 components:
They're caused by dirt on the front element or filter of the lens. The images vary in size
depending on focus, so they appear (sharply) at a different place on each component. It
seems to be a characteristic of the “PMax” algorithm that they're not optimized out. “DMax”
(first image) does better, but it doesn't give the same depth of field (run the cursor over an image to compare it with its neighbour):
Up early this morning for a busy day. First, the arrival of Mick Solly, the gardener. A
good thing I did, too: almost immediately I got a call from his wife, Eileen, telling me
that he was sick, and wouldn't be able to make it until next
week.
The other thing on hand was a trip
to Ballarat to hear the results of
Yvonne's pancreas scan, almost exactly a
year since the last one. Another phone call: Mr. Shimokawa was tied up in an
operation and wouldn't be able to make it. Postponed until Monday.
My web site load has dropped from the excessive load averages I have seen yesterday, but
only to about 18. Why such a high load? Previously I had guessed at the multiple names for
the web server, but it's been a
week since I've been sending permanent redirects. Still, there are a number of
requests coming in to oldwww. Played around with the web server configuration there
to log redirects differently depending on the server name. That helped. After a few
minutes my log files showed:
-rw-r--r-- 1 root wheel 0 Oct 18 06:41 lemis-access.log
-rw-r--r-- 1 root wheel 3502 Oct 18 06:45 mail.lemis-access.log
-rw-r--r-- 1 root wheel 52586 Oct 19 01:10 people.lemis-access.log
That was surprising. I had expected lemis-access.log to be the biggest. And who's
talking to mail.lemis.com? Almost only semrush.com! It must have gone out and tried all known names to even discover
that mail.lemis.com had a web server behind it. And now it seems that it has ignored
all the 301s and continues happily. You'd expect better from a company that sells web
services. semrush.com is also well represented on people.lemis.com, but there
were a lot of others too, a large number with no reverse lookup. I wonder why I even
created the alias.
Still, they've had plenty of warning. Why not just shut the server down? Bingo! The load
average dropped to round 4. Maybe I don't need a bigger server after all.
I had put the remaining Thai basil and the chili plant on the table outside the lounge room
until I could find a good place to plant it. The basil will probably be given away, but
what about the chili? The weather forced my hand: the wind blew the pot off the table,
completely uprooting it. OK, in front of the north window of the lounge room:
So I left the old web server stopped overnight, so that it didn't redirect requests to inappropriate
domain names
to the new server. Came back in the morning and the load
average on the new server was still below 10. Turned on the old server again, and bang!
Within minutes I had:
last pid: 16293; load averages: 31.65, 23.93, 13.76 up 53+19:32:08 01:16:42
Why does it make such a difference? Is it the content that
keeps PHP busy? Sadly, it doesn't seem to
have had much effect on the traffic.
Both are only marginally sharp, probably because of the wind. They're acceptable at the
standard (“tiny”) size, but even a slight enlargement shows how bad they are:
That's a lot better! But there are still details. I did this with Zerene stacker, which offers two different algorithms.
I must really read the details, but so far I just have the names: PMax (apparently
preferred) and DMap. The first photo came out better with DMap. The PMax version had a
halo round the features. Here DMap, then PMax (run the cursor over an image to compare it with its neighbour):
Camera lenses tend to last longer than the bodies, and even in the digital age I have a
couple of lenses that I have had for over 10 years. But then there are the analogue lenses.
I've had my current 50 mm f/1.4
Super-Takumar for over 52 years, and some of the lenses I have are clearly older than
that.
The 50/1.4 Super Takumar had an excellent reputation, but three years ago I received my Leica Summilux 25 mm f/1.4 and discovered that it's an order of magnitude better than
the Super Takumar:
The Super Takumar is designed for 35 mm film (24x36 mm). The camera I was using has a
sensor only half that size. In other words, these “corner” shots are nothing of the
kind: they're halfway between the centre and the corner. At the corner it must be even
worse.
There are lenses which even in those days had a poor reputation. My very first real
camera, the Voss
Diaxette, came with a 45 mm f/2.8 Steinheil Cassar,
a Cooke triplet! That's a lens
design that became obsolete in 1902 with the advent of
the Tessar. It was very soft
at the edges at full aperture. By chance I now have a new Cassar, 50 mm f/2.8, which
came with my newest Edixa
Reflex.
So: how about comparing the Super Takumar and the Cassar? Certainly the appearance is
markedly different:
To do it right I need to find a “full frame” camera that will take Pentax screw thread
lenses, clearly with an adapter. In the meantime it would be interesting to see how it did
on the Olympus OM-D
E-M1 Mark II.
The results are only partially of interest, since the camera is wrong. The big surprise was
the difference in contrast:
Why that? The “stars” in the background from the Cassar are almost invisible, and the
foreground is washed out. Why? Poor quality antireflection coating? But the coma isn't
significantly worse than on the Super Takumar. For that I'll have to find a “full frame”
camera.
This bush is in Stones Road, on the way from our house to the “schoolyard” (the junction
with Bliss Road). I've seen it flower before, but I think these are the first photos of the
buds:
The kitchen shower head that I ordered on eBay have arrived. One was fine, but the other had a sticky jet/spray switch, allowing it to
produce both at the same time: