Recent searches
Search options
#sysadmin
#Realtoxmedia as a hoster started out great: 2tb/40gb SSD/4GB RAM for 5€/month? Is really good and has been working fine, but recently, the HDD array maxed out at 2-4mb/s read, though 100+mb/s were possible a month ago or so. Let's see how they'll handle it. Because I made it pretty clear that I'm pissed.
Question for people who use #rkhunter on #FreeBSD:
Wait, first a prelude: The out-of-the-box defaults are so nice, just turn it on in periodical and you get daily reports that are actually meaningful and easy to understand.
Okay,the question: Given that it's still extremely verbose, pages and pages and pages of checks that have passed, what does an alert look like?
Do I have to scroll down all the way to read the exact same unchanging wall of text every day, on every host, for months and months, to have any hope of spotting the word "Failed" on line 2760 one day? Or does it make a bit of noise, put something in all caps at the top of the email, something that I'm likely to notice through the banner blindness?
After I tested for system stability I started thinking on how to prevent further disk failures and when it happens how to make it less painful to restore.
Full disk cloning.
I'm going to clone my entire disk quite regularly along with my existing timeshift, borg backups.
I used Clonezilla in memory mode to take my first full disk clone back up and stored it in my homelab's hard disk.
contd.
And it was definitely not easy!
I rsynced timeshift to root and struggled couple of days to restore /home from the Borg mount as I messed up the mount points.
I changed the UUIDs of the drives in the fstab, crypttab and reinstalled grub; but for some reason the grub kept looking for old UUID of the Luks root partition.
I had to manually edit the UUID to unlock and fix the issue by reinstalling kernel/grub. Had to restore the SELinux contexts as well to boot.
contd.
Perhaps the heat sink in Blade S70 made it last for 2 years? Should I consider buying a 2TB version of the same? I decided not as ADATA seems to be replacing controllers in a whim for different regions and I'm quite sure they cheaped it out in India like everyone does.
Perhaps Blade S70 needed a heat sink because the controller demanded it. So I went with Kioxia Plus G3 itself.
Now I have to restore /root from TimeShift and /home from Borg.
contd.
Actually my new SSD borked within a week, I was so stressed that I didn't even post about it. I have gotten over the PTSD that I will post about it today. Those who don't know the BG, please read my earlier posts.
TL;DR my 1TB OS drive showed signs of failure and I recovered my data using various strategies and replaced the the drive with 2TB WD Blue SN580.
Now the new drive failed within a week after I ran 70B LLM model.
contd.
Time to #introduce myself here at m.h-town.
I'm a #sysadmin with more decades in service than I care to think about. I'm a Unix guy who deals with Linux when I have to.
I am heavily into musical theatre ( #musicaltheater ) but don't act/perform. I mostly do lighting - most recently as Master SpotOp for Seattle Musical Theatre.
I love to cook.
I'm gay, single, and live in Seattle.
I generally won't accept follow requests from blank profiles.
Any other questions? Just ask.
I send out magic waves on the wires. A remote machine wakes up.
Not sure if I asked this before: Does anyone use anything in particular to inject #apache logs into #SQL databases? I have been looking around and asking around and the only solid I got was "do not expect an apache module for that; it would introduce too much latency to each request" in #httpd@libera.chat.
Backups are only good as long as they work.
Spent this morning, restoring the backup of our Mastodon instance burningboard.net to a fresh virtual machine in Proxmox and did some validation that it is valid, complete, restorable and the disaster recovery documentation is up to date.
Everything worked perfectly 
Next restore-test: 10/2025