WTF Bites
-
The "I Hate Oracle Club"
threadcategory is .
-
I was installing microk8s for testing purposes on my work notebook. After enabling a bunch of things, specifically dns, ingress, rbac and registry, the kubelet started logging swaths of errors like
Apr 21 19:41:20 ntb1234567 microk8s.daemon-kubelet[12388]: E0421 19:41:20.527281 12388 reflector.go:178] object-"kube-system"/"coredns-token-p7gn4": Failed to list *v1.Secret: secrets "coredns-token-p7gn4" is forbidden: User "system:node:NTB1234567" cannot list resource "secrets" in API group "" in the namespace "kube-system": no relationship found between node "NTB1234567" and this object
and the pods were not starting.
Well, the notebook had its hostname set to
NTB1234567
, which is its inventory number.1: Like that, with uppercase prefix. Yes, DNS is case-insensitive, nothing should care, right?
2: Something lowercased the name when setting the name of the node, but not any of the other places including the service account.Changing the name to lowercase and updating all the configs it generated fixed it.
-
Microsoft AutoUpdate
All Apps are up-to-date Check for updates
ο Automatically keep Microsoft Apps updated Settings...Um, ok, good. Thanks for this information?!
Why the hell are you popping this up while I'm busy doing other things? Does "automatic" mean "but I need a pat on the head every once in a while"??Since I posted this ("6 days ago") I got that message two more times.
Yesterday it wasn't quite as bored, apparently, and instead presented me this:Microsoft AutoUpdate
Updates available: 1 Update
ο Automatically keep Microsoft Apps updated Settings...Do the people at Redmond have a different definition of "automatic" than normal people?!
-
@boomzilla said in WTF Bites:
Implementation details. 4000 is the max size for a varchar2 field (unless you do some non-standard DB configuration, which we haven't).
It's been a little while since I worked with Oracle DB but I seem to recall that you can do
varchar2(4000 char)
instead of4000 byte
for all your fancy foreign spaces and Polish Δ 's with dingleberriesThere is still a 4000 byte limit unless you change the database configuration.
-
Do the people at Redmond have a different definition of "automatic" than normal people?!
And they also have a different definition of "updates"
-
If Amazon has worked out time travel then why don't they use it to stop that asshole from eating the bat?
-
@Polygeekery said in WTF Bites:
If Amazon has worked out time travel then why don't they use it to stop that asshole from eating the bat?
Who says they would try to stop it?
-
@JBert ......good point.
Maybe they shipped the bat back in time to Wuhan so that they could make millions of people dependent on them for toilet paper?
-
@Polygeekery Who says that's May 3rd 2020?
-
@Polygeekery said in WTF Bites:
@JBert ......good point.
Maybe they shipped the bat back in time to Wuhan so that they could make millions of people dependent on them for toilet paper?
If there's anyone profiting of this shit show, it is indeed Amazon.
-
@Polygeekery said in WTF Bites:
@JBert ......good point.
Maybe they shipped the bat back in time to Wuhan so that they could make millions of people dependent on them for toilet paper?
If there's anyone profiting of this shit show, it is indeed Amazon.
Not only. This is a live security feed from Charmin headquarters:
-
@hungrier Ah yes, the literal shit show.
-
Not only. This is a live security feed from Charmin headquarters:
I recently had delivery of this tshirt.
-
@Polygeekery said in WTF Bites:
If Amazon has worked out time travel then why don't they use it to stop that asshole from eating the bat?
It only works a maximum of 4 days, unfortunately.
-
@Polygeekery Actually, you have to return it by May 5. They can deliver it by May 3 because that's just after the previous dude has to return it.
-
Actually, you have to return it by May 5. They can deliver it by May 3 because that's just
afterbefore the previous dude has to return it.
-
-
WTF of my day: So, turned out that this Logitech Webcam I had lying around ... lied about its condition. Because, y'see, a webcam's working condition would, by necessity, included "producing a moving picture in real time".
This one turned on the green light, captured audio but otherwise produced a wonderful full-black stream.
A short Google search yielded other people with the same webcam and the same problem. Seems that Logitech didn't update the drivers to work properly in later releases of Win10.
Hrmh. That's when I remembered that I had this KinectV2 from my XboxOne. And I also remembered that Microsoft had actually provided software and drivers for it to work on PC - surely that one can also double as a webcam?
Whipped up an adapter (why on Earth did they plonk a 12V powerline directly onto the USB3 connector?) and off we went! Plugged it in, the lights went on, drivers installed and ... nothing.
Windows10 then insisted that there was no camera. Downloaded the Kinect2.0 SDK - well, the gizmo works, I can get a wonderful 3D point cloud of myself sitting in front of the PC.
But still nothing that would actually be useful in a regular meeting.
Further research yielded something called a KinectCam. Installed that - still nothing (well, it was written for Win8). Another guy had updated the sources for KinectCam but it could only compile to a 64bit executable now, but the included installer needed 32bit...
One last ditch effort: Device Manager -> Kinect Sensor Device -> Rightclick on WDF Kinect Sensor -> Update driver
And what do you know? It works now. I can even use Windows Hello to logon to my PC now.
-
@Kamil-Podlesak said in WTF Bites:
I think that even in Oracle, you can treat CLOBs just like big VARCHARs these days
Is an empty CLOB null?
-
@Kamil-Podlesak said in WTF Bites:
I think that even in Oracle, you can treat CLOBs just like big VARCHARs these days
Is an empty CLOB null?
An empty CLOB is saved as null.
However, looking for an empty string will not match CLOBs with null values.
You CAN set the CLOB to an empty value by setting it to EMPTY_CLOB() instead, should you feel like it. It still won't match a where CLOB LIKE ''...
-
https://status.snapcraft.io/ says:
Note the reverse chronological order.
-
@Rhywden I've had my own odyssey with (web)cams in the past few weeks. I'm not too fond of them to begin with, which means that I never bought one for my desktop at home.
For events that required video, I initially started using a laptop and its potato cam. Balancing the laptop on an ad-hoc collection of stuff on my desk, as to position the camera in a useful place, got old pretty quickly. Besides desk-space and such annoyances it makes stuff like briefly sharing a window from the screen a bit messy, because they tend to preceded by a "one sec, need to switch computers for a moment" (or, alternatively, repeated clarifications of why there are two of me).
Of course, webcams are kinda sold out all over the place, at least the ones that I would consider being worth getting. (On one hand, I can already get potatoquality with the laptop. On the other hand, I'm also not going to spend several hundred β¬β¬β¬ on a fucking webcam.)
Somebody suggested using a smartphone. There are apparently apps for that. Consequently, the current solution involves an old smartphone with a smashed screen taped to a monitor, running some random android app that streams the camera over the wifi connection. I can receive the video stream on my computer and turn it into something that looks like a webcam thanks to the magic of v4l2loopback devices.
On the plus-side, I can fuck with the video stream in the middle. I'm considering plugging in a face detector and making my eyes glow. Or something. (Or perhaps have it detect when I'm about to nod off, and then quickly switch the stream to a prerecorded one.)
On the down-side, most android apps that I've found suck ass. Most transmit MJPEG videos, which combine the qualities of overcompressed JPEG images with those of a slideshow. There are a few that can stream h264, though. They have their own issues, but so far those seem tolerable. The smartphone reaches meltastic temperatures, though.
-
So do I continue or continue or continue or...
Each continue is a different combination of those 3 APIs. The continue button that's greyed out is the one with none of them selected (so you can't continue). You see all of them because their CSS is broken.
Hey, it seems plausible, at least. There's 8 buttons... 2^3 = 8.
-
says:
Note the reverse chronological order.
Well... It makes some kind of sense if you consider that you generally read top to bottom and want to find the important issues ("api is broken", "api trouble is fixed") as soon as possible.
-
@JBert The reverse chronological order does make sense. The fact that the issue was fixed first and then reported over and overβ¦
-
Sometimes, the Ubuntu launcher sidebar forgets that Postman is open
-
@JBert The reverse chronological order does make sense. The fact that the issue was fixed first and then reported over and overβ¦
There was no issue
-
@hungrier Unrelated:
$ whereis docker-compose docker-compose: /usr/local/bin/docker-compose $ docker-compose up -d [service] bash: /usr/bin/docker-compose: No such file or directory
Also : in another terminal tab, it works no problem. A couple hours ago, it was working everywhere.
-
@hungrier sounds like hashed paths.
-
@Bulb (after looking up hashed paths)
How did it get hashed to the wrong thing though? Is my work Linux VM secretly smoking hash?
-
@Bulb (after looking up hashed paths)
How did it get hashed to the wrong thing though? Is my work Linux VM secretly smoking hash?
You're thinking too much about it man, just take another toke.
-
@hungrier Could some update have happened meanwhile (perhaps automatically; there is unattended-upgrades these days) and installed the
/usr/bin/docker-compose
with an incorrect interpreter? Because that error usually means the file has incorrect value after the#!
at the beginning.
-
@Bulb I think I've figured it out. Yesterday, I updated docker-compose, and for I had to remove the old version and reinstall it using pip. But I guess I did that and all my subsequent work in my other tabs, and left that original one alone until just now. And apparently the old version was in usr/bin instead of usr/local/bin, and that's where that wrong path came from.
-
@hungrier Well, if the old version was one from a package, it is completely expected.
/usr
is for things from packages (deb, rpm),/usr/local
is for more manual ways like pip. I think some old pips may have installed in/usr
for (something about python paths) though.
-
@Bulb That makes perfect sense (except for which is that the
apt
version of docker-compose only goes up to an old version that didn't have the feature I needed)
-
except for which is that the apt version of docker-compose only goes up to an old version
It's a problem with the other parts of docker too, not just compose. CentOS had even more ancient versions.
Also, Debian only has docker-compose until stretch (a very ancient 1.8) and then in sid, but not even in current testing (bullseye).
-
The whole container ecosystem is in kind of a mess.
- RedHat has already fully switched to cri-o + podman, but it had a pretty severe bug with piping input into the container until very recently, which would have rendered it useless if I got around to trying it.
- I wanted to try podman because it runs under the user launching it and utilizes subusers instead of posting everything to a system-wide daemon like docker, but the atomic-project ppa is dead while official Debian packages are stuck in the new-and-byhand queue since 14th january and already outdated.
- Docker-compose is an alien when the rest of your world is centred around Kubernetes manifests, but for some use-cases it's all you've got.
- Podman can read Kubernetes pod manifests, but they can do 1% of what docker-compose can (no environment substitutions, no override files; kompose would be great, but I doubt it's anywhere close)
- And then there is more mess in Kubernetes land. There is helm, which has great capabilities like rollbacks, including automatic on failure to start, but the composition is kind of half-baked and using go (~moustache) templates for yaml is β¦ weird at best.
- There is kompose, which was already integrated into kubectl itself, which provides the right kind of flexibility, but then it does not have the history recording and rollbacks and tests and all that other good stuff.
Yeah, everybody rushes ahead with half-baked ideas, nothing ever stabilizes and distributions can't keep up.
-
The whole container ecosystem is in kind of a mess.
- RedHat has already fully switched to cri-o + podman, but it had a pretty severe bug with piping input into the container until very recently, which would have rendered it useless if I got around to trying it.
- I wanted to try podman because it runs under the user launching it and utilizes subusers instead of posting everything to a system-wide daemon like docker, but the atomic-project ppa is dead while official Debian packages are stuck in the new-and-byhand queue since 14th january and already outdated.
- Docker-compose is an alien when the rest of your world is centred around Kubernetes manifests, but for some use-cases it's all you've got.
- Podman can read Kubernetes pod manifests, but they can do 1% of what docker-compose can (no environment substitutions, no override files; kompose would be great, but I doubt it's anywhere close)
- And then there is more mess in Kubernetes land. There is helm, which has great capabilities like rollbacks, including automatic on failure to start, but the composition is kind of half-baked and using go (~moustache) templates for yaml is β¦ weird at best.
- There is kompose, which was already integrated into kubectl itself, which provides the right kind of flexibility, but then it does not have the history recording and rollbacks and tests and all that other good stuff.
Yeah, everybody rushes ahead with half-baked ideas, nothing ever stabilizes and distributions can't keep up.
It's like web devs started doing backend stuff.
-
-
It's like web devs started doing backend stuff.
-
@Applied-Mediocrity I can't tell whether the WTF is the window borders overlapping the Project Location dialog or that you have a gazillion "Local Disk"s.
Edit: Or that you actually called your Windows user account "Root".
-
@TwelveBaud said in WTF Bites:
gazillion "Local Disk"s.
Yeah! At least label them, like
sda1
sde2
, etc....
-
@TwelveBaud Yes
-
-
I just got a gold nugget from the information service projekt...
"We don't check that requests get the correct response. You have to do that yourselves."
How the fuck do you even get request/response wrong so bad that people get someone elses responses?
-
I just got a gold nugget from the information service projekt...
"We don't check that requests get the correct response. You have to do that yourselves."
How the fuck do you even get request/response wrong so bad that people get someone elses responses?Probably by creating a really simple FIFO queue based on
Math.random()
when deciding who gets what?
-
I just got a gold nugget from the information service projekt...
"We don't check that requests get the correct response. You have to do that yourselves."
How the fuck do you even get request/response wrong so bad that people get someone elses responses?Probably by creating a really simple FIFO queue?
Well, this is SOAP, so simple is already dead and buried. Of course, that doesn't prevent someone from doing the derpful.
-
@Carnage Well, I meant "simple" more in the "village idiot" sense.
-
@Carnage Well, I meant "simple" more in the "village idiot" sense.
Yeah, and taking that they said they don't guarantee that the response ends up at the correct recipient, village idiotry is guaranteed here.
-
The whole container ecosystem is in kind of a mess.
- RedHat has already fully switched to cri-o + podman, but it had a pretty severe bug with piping input into the container until very recently, which would have rendered it useless if I got around to trying it.
- I wanted to try podman because it runs under the user launching it and utilizes subusers instead of posting everything to a system-wide daemon like docker, but the atomic-project ppa is dead while official Debian packages are stuck in the new-and-byhand queue since 14th january and already outdated.
- Docker-compose is an alien when the rest of your world is centred around Kubernetes manifests, but for some use-cases it's all you've got.
- Podman can read Kubernetes pod manifests, but they can do 1% of what docker-compose can (no environment substitutions, no override files; kompose would be great, but I doubt it's anywhere close)
- And then there is more mess in Kubernetes land. There is helm, which has great capabilities like rollbacks, including automatic on failure to start, but the composition is kind of half-baked and using go (~moustache) templates for yaml is β¦ weird at best.
- There is kompose, which was already integrated into kubectl itself, which provides the right kind of flexibility, but then it does not have the history recording and rollbacks and tests and all that other good stuff.
Yeah, everybody rushes ahead with half-baked ideas, nothing ever stabilizes and distributions can't keep up.
Is this the cricket thread?
I understood approximately zero of those words.