|
xzzy posted:How big is the database? Is anything running some potentially goofy queries against it? In my experience such queries reveal themselves by filling up /tmp first but maybe your config is a bit different with a relatively big /tmp. Various databases on the server which vary in size. Biggest is about 200G, total is about 500G on the server. /tmp is 20G although we configure temporary tables into a specific database rather than going to /tmp, so most things go there. Disk is 750Gish - we've had space cause problems with /tmp filling up before, but it's not he issue here at the moment. Someone is almost certainly running goofy queries against it - that's a different battle to fight but "It's always worked before" means this is my problem not theirs for the moment.
|
# ? Dec 15, 2020 11:11 |
|
|
# ? May 29, 2024 15:33 |
Not to go off on a tanned gent too much, but do people still choose MariaDB over postgres on new deployments?
|
|
# ? Dec 15, 2020 12:16 |
|
BlankSystemDaemon posted:Not to go off on a tanned gent too much, but do people still choose MariaDB over postgres on new deployments? unfortunately bad software doesn't support postgres
|
# ? Dec 15, 2020 12:26 |
|
BlankSystemDaemon posted:Not to go off on a tanned gent too much, but do people still choose MariaDB over postgres on new deployments? There are still people spinning up new LAMP stacks every day. If you've been writing cake/php apps against mysql for 15+ years, I'd imagine it would be hard to switch tech at this point I personally haven't heard of anyone using anything but postgres in the last 10 years. Postgres 9.4 (2012?) was a really fantastic release, 9.6 got native json support, 10.0(?) got logical replication and now they're up to v13.something and it's running pretty much every startup company at this point Pre-containers maybe the hassle of installing/configuring the db from scratch was a reason to pick one over the other but now it's a literal one liner to "install postgres" as a container
|
# ? Dec 15, 2020 12:33 |
|
Major Ryan posted:I don't know to a more specific databases thread, so if anyone here could help, it'd really be appreciated since this is doing my head in. Do you have a web server running on there, or is it just a database server? Show some top output Post your my.cnf Post your mysql.err You'd actually want to make the buffer_pool_size smaller if you're running out of memory. Usually when I've seen this happen, it's because of some query that's returning a billion rows of data. And usually chucking another 32GB of memory in the server isn't very expensive.
|
# ? Dec 15, 2020 13:52 |
|
Bob Morales posted:Do you have a web server running on there, or is it just a database server? I'll have to grab some bits. Top is fine at the moment because it's not busy right now, so if I can catch something when it crashes next. It's just the db server, there's nothing else on here apart from the OS basics and the Maria service. We've been tweaking the buffer_pool_size for a while, it would have been 24G, then 15G, 10G, now we've tried 5G... More memory is an option and we can do that to make the problem go away (which is probably the answer we'll take to be honest). I'm just confused from a technical point about the buffer_pool_size - it's obviously not the hard memory usage limit that I thought it was. my.cnf posted:[mysqld]
|
# ? Dec 15, 2020 14:29 |
|
Major Ryan posted:More memory is an option and we can do that to make the problem go away (which is probably the answer we'll take to be honest). I'm just confused from a technical point about the buffer_pool_size - it's obviously not the hard memory usage limit that I thought it was. There are a million other things that mysql will use memory for, the buffer pool size just limits one thing https://www.percona.com/blog/2018/06/28/what-to-do-when-mysql-runs-out-of-memory-troubleshooting-guide/ Are you monitoring all your mysql stats with something? You can get a lot of metrics out of it.
|
# ? Dec 15, 2020 15:04 |
|
Bob Morales posted:There are a million other things that mysql will use memory for, the buffer pool size just limits one thing Yeah, the gap isn't in the monitoring, it's in the lack of understanding of what it means :-) There's someone with more SQL knowledge that me looking into it too, so hopefully they'll be able to offer some suggestions based off the monitoring. The user-stats line in the my.cnf is for the percona monitoring and management tools that we've just set up, so we've got some nice grafana graphs to go through and try to spot something. Thanks for the help; it's enough to know that the buffer pool size isn't the limit I thought it was. At least confirms there's memory usage somewhere else in Maria, just need to find out where...!
|
# ? Dec 15, 2020 16:43 |
|
I'd log queries (especially slow ones) to see what is happening before the kernel goes on a killing spree. Without knowing what it's actually doing it will be hard to figure out the problem. It could be creating in-memory tables and filling them or something similarly crazy.
|
# ? Dec 15, 2020 18:55 |
|
Major Ryan posted:Yeah, the gap isn't in the monitoring, it's in the lack of understanding of what it means :-) It means the same thing every sysadmin problem means: users are always the problem and this job would be so much easier if they stopped using their computers so drat much. Or in this case quit running queries that result in enormous joins.
|
# ? Dec 15, 2020 19:06 |
|
Bob Morales posted:Does it not increment the .1 because you have the date in the file which changes for each file? That's correct, the real file names increment by date. This is handled by syslog-ng automatically at the start of each day (midnight). Below is an example of actual filenames from my systems. code:
xzzy posted:logrotate has limitations dealing with files that have different file names. If you match them with a glob it runs the config against each match as a separate task, not as different versions of the same log file. Thanks for the advice, I had read a similar response on Stack Overflow. I just wanted to confirm with the masses before I start looking for a new method to rotate these files. I'm honestly looking to keep 7 days worth of log files on these systems. Anything else can be thrown away. Any advice on the specific approach I should use? From my perspective I could: 1) Create a cronjob in cron.daily and leverage the find command to search for logs older than 7 days, then delete. I don't like managing cron jobs though. 2) Create a hacky log-rotate script that leverages the post-rotate method, to call the find command, search for logs older than 7 days, then delete. 3) Use syslog-ng natively to handle deletion of old log files. (Not even sure if this is possible TBH).
|
# ? Dec 15, 2020 21:16 |
|
I think I used to have a setup where either logrotate or something else compatible with it was rotating logs like that for me. But I don't remember how those settings worked. It might just be an exotic option in the logrotate config.
|
# ? Dec 15, 2020 21:31 |
|
Unless the business logic required dates in the filename, I'd alter my syslog config to name the log "stacktrace.log" and use the mtime of the files if someone wanted to know what day the log was from. Logrotate has an option to append the date if it's really a requirement (dateext I think, but it has a few "date" options).
|
# ? Dec 15, 2020 22:00 |
|
I would just include the date in the entry in the file
|
# ? Dec 15, 2020 22:59 |
|
Reading the release announcement for the next OpenSuSe leap:quote:openSUSE Leap 15.3 is based on the Jump concept that was developed over the past several months, which makes it and SUSE Linux Enterprise compatible. openSUSE Leap aligns with SLE and its Service Packs (SP), which keeps the system updated, stable and patched. Upon General Availability of this release, there will be a whole new level of harmony between Leap 15.3 and SUSE Linux Enterprise 15 SP3. So they are doing literally the opposite of RH and appear to be kind of smug about it. I guess I'll try it out anyway, I kind of want to play more with OBS.
|
# ? Dec 16, 2020 16:16 |
|
I've brought up opensuse in my work group as a "haha just kidding but not really, anyone wanna try it?" when discussing the centos stuff and there's zero interest. Users are the same way, all they care about is which free redhat derivative will give them the longest shelf life between updates. Which makes no sense to me, if you're so angry at redhat for killing your distro that you want to jump ship from centos, why the gently caress would you continue to use anything redhat based? What stops them from gobbling up the "new savior" Rocky Linux in a few years and killing it too? I don't understand people.
|
# ? Dec 16, 2020 16:23 |
|
xzzy posted:I've brought up opensuse in my work group as a "haha just kidding but not really, anyone wanna try it?" when discussing the centos stuff and there's zero interest. Users are the same way, all they care about is which free redhat derivative will give them the longest shelf life between updates. Non linux people don't realize that redhat != Linux, so if you suggest a different distro they react as if was a change to AIX or BSD or some unix that is arcane and bizzare. This includes alot of windows IT people
|
# ? Dec 16, 2020 16:33 |
|
I just remembered that Fedora Server exists. Who are its target users, exactly? People who run pet servers and can afford to update them both carefully and frequently? People who for some reason need to keep package versions synchronized between their workstation(s) and their server(s)?
|
# ? Dec 16, 2020 17:28 |
|
NihilCredo posted:I just remembered that Fedora Server exists.
|
# ? Dec 16, 2020 17:30 |
|
NihilCredo posted:I just remembered that Fedora Server exists. When I worked for a telecom in their attempt at a cloud solution, they used a customized fedora 21 as their server OS for running xen on top of, so them?
|
# ? Dec 16, 2020 17:31 |
|
NihilCredo posted:I just remembered that Fedora Server exists. i decided i wasn't larping an enterprise setup at home with regards to uptime percentage or support. it works great, i don't know what you mean by updating carefully and frequently
|
# ? Dec 16, 2020 17:50 |
|
NihilCredo posted:I just remembered that Fedora Server exists. According to the IRC channel like 20% of their downloads are server
|
# ? Dec 16, 2020 18:03 |
|
Is it also 20% of their ISOs because that's probably just mirrors or something, unlike actual usage data
|
# ? Dec 16, 2020 18:11 |
|
hifi posted:i decided i wasn't larping an enterprise setup at home with regards to uptime percentage or support. it works great, i don't know what you mean by updating carefully and frequently My enterprise larp is flatcar, gently caress the OS, put it all in containers. I've been really happy with it, if I want to try something new pull the image. When I decide I hate it, delete it and it was never there. Never get that impulse to reinstall because I got too many extra packages or dead config installed.
|
# ? Dec 16, 2020 18:28 |
|
I use Fedora Server in VMs/for messing with poo poo, because yes I use Fedora Workstation and it's easier. Nothing production-ish but it would probably be fine enough . I do plan on switching my docker host to Fedora CoreOS once it's a little more baked.
|
# ? Dec 16, 2020 18:48 |
|
xzzy posted:I've brought up opensuse in my work group as a "haha just kidding but not really, anyone wanna try it?" when discussing the centos stuff and there's zero interest. Users are the same way, all they care about is which free redhat derivative will give them the longest shelf life between updates. I'd say it's mainly familiarity and software compatibility. Apt, yum and zypper are very similar, but still annoyingly different from each other, not to mention any other components. Especially if you want to do automated installs. And sometimes you need to install software outside the distribution and you better hope it's packaged for the distro you are using. Red Hat has always been a popular packaging target by upstream and you were almost certain it would work in CentOS. And then there is the research group which wants a new server and gives a list of software they need. Software A is packaged for Red Hat by upstream, software B for Ubuntu. Good luck choosing the distro. It's very trivial to prevent the same from happening to Rocky Linux. Don't give Red Hat the trademark. The only reason Red Hat could make this announcement is because they own the CentOS trademark, and they own it because the people behind CentOS donated the trademark with Red Hat took it under their umbrella.
|
# ? Dec 16, 2020 19:06 |
|
I still don’t understand why they don’t standardize package managers. I mean I only use Debian based distros so far but it’s odd that there are so many variants to do the exact same thing. gently caress snap though.
|
# ? Dec 18, 2020 07:04 |
|
SpaceSDoorGunner posted:I still don’t understand why they don’t standardize package managers. I mean I only use Debian based distros so far but it’s odd that there are so many variants to do the exact same thing. who is "they"? package managers tend to be the same or very similar within distro families, tho the gui may vary wildly. apt is apt, yum is yum, pacman is pacman one reason I always do my updates from the cli
|
# ? Dec 18, 2020 07:09 |
|
RFC2324 posted:who is "they"? package managers tend to be the same or very similar within distro families, tho the gui may vary wildly. apt is apt, yum is yum, pacman is pacman Well, KDE uses discover (I think) and I think Gnome uses something else. I just use the CLI, because that always just works.
|
# ? Dec 18, 2020 07:42 |
It's most likely a combination of NIH syndrome and insufficient knowledge about other options that're available.
|
|
# ? Dec 18, 2020 12:24 |
|
What’s the least headache inducing GUI for base Debian? I’ve tried Gnome and Xfce and both has the disappearing desktop/taskbar issue that I’ve seen referenced online and in books and had to rely on using the win key to pull up anything from the GUI. I know Kali and Kali variants I’ve used use xfce and I never have this problem. Any good fixes? One of my projects was to create an up to date buscador off of base Debian rather than use Ubuntu like Bazzell recommends now for ease of use. I tried a month ago and I ended up giving up just because there were too many small headaches but it did teach me a lot about Linux.
|
# ? Dec 18, 2020 14:10 |
|
never not KDE e: this goes for any distro not just debian, really. out of the box, kde acts more or less like a good version of windows would, and pretty much everything can be configured to work the way you want Truga fucked around with this message at 14:31 on Dec 18, 2020 |
# ? Dec 18, 2020 14:21 |
|
SpaceSDoorGunner posted:What’s the least headache inducing GUI for base Debian? I’ve tried Gnome and Xfce and both has the disappearing desktop/taskbar issue that I’ve seen referenced online and in books and had to rely on using the win key to pull up anything from the GUI. This is a personal choice, I would always keep using Xfce
|
# ? Dec 18, 2020 15:42 |
|
Truga posted:never not KDE I'm not a debian person, but KDE is always thr answer, unless the question is "which de do I install on this potato?"
|
# ? Dec 18, 2020 16:35 |
|
I always end up having to install both because users are very fickle. (mostly on vnc servers, which is about the only case where you have multiple people running desktops on one machine)
|
# ? Dec 18, 2020 16:36 |
|
I never, ever, ever use the gui package managers anymore. In my experience they always gently caress up eventually.
|
# ? Dec 18, 2020 17:54 |
|
VostokProgram posted:I never, ever, ever use the gui package managers anymore. In my experience they always gently caress up eventually. I stopped using them years ago because of a habit of loving up the package database, making the whole system fucky. The worst case left me only able to install tarballs the old fashioned way because the rpm database had so completely poo poo itself
|
# ? Dec 18, 2020 18:03 |
|
VostokProgram posted:I never, ever, ever use the gui package managers anymore. In my experience they always gently caress up eventually. So do the command line ones. On our compute servers that are running full tilt 24/7 yum corrupts the rpmdb on steady percentage of nodes every week. It's single digit percentages, but still enough to be annoying.
|
# ? Dec 18, 2020 18:04 |
|
xzzy posted:So do the command line ones. On our compute servers that are running full tilt 24/7 yum corrupts the rpmdb on steady percentage of nodes every week. It's single digit percentages, but still enough to be annoying. It never corrupts the rpmdb badly enough to be unrecoverable tho, right? My experience with cli utils is that when they do gently caress up you can use the backend to them(rpm or whatever the equivalent is on deb systems) to fix the problem. Mildly annoying vs requiring that you flatten and reinstall. Everything fucks up eventually on linux
|
# ? Dec 18, 2020 18:09 |
|
|
# ? May 29, 2024 15:33 |
|
Yeah, rpm --rebuilddb fixes it. We never should have invented computers.
|
# ? Dec 18, 2020 18:22 |