@brot@social.tchncs.de

sysadmin

This profile is from a federated server and may be incomplete. Browse more on the original instance.

milan, to random
@milan@social.tchncs.de avatar

are there sata->usb adapters with full smart support (i mean i think there are but i am not sure how to look them up) or is this a limitation of my 2.5'' oem harddrives? also the transcend thing i have only reports 0°C status ok, no error logging support

brot,
@brot@social.tchncs.de avatar

@milan those with usb attached scsi support should allow to access the smart status of the drives.
Since I upgraded to proper controllers with UAS support, I had no problems backing up to a remote raspberry pi 4 with btrfs raid 1.
(Even with the disks LUKS-encrypted and having zstd encryption active, rsync reports rates up to 50 MB/s. Modern SoCs are kinda magic)

milan, to random
@milan@social.tchncs.de avatar

great, this host suffers from https://github.com/openzfs/zfs/issues/14005 again, after i was sure i had it resolved and dont know whats different here from other hosts... maybe i should use btrfs more.

brot,
@brot@social.tchncs.de avatar

@milan btrfs serves the files on my personal nas since... 8 years at lower, and has survived a drive slowly dying twice. And on my personal laptop it has been the rootfs since 2011.
Early problems have been lower performance, but that is an area where there has been constant improvement in the last few years and it is good now.
I can't vouch for raid5 or 6, but other than that it has done it's job without being annoying :)

milan, to random
@milan@social.tchncs.de avatar

after i had to copy files to 40 cheap usb2.0 thumbdrives using rsync ... and sometimes multiple times as some files somehow got damaged, even with using sync command and checksum option and stuff ... i am never trusting that intenso brand again. just had to share a cloudlink as at least one person had almost all files damaged (not everybody got theirs yet).

brot,
@brot@social.tchncs.de avatar

@milan one thing you could have tried is to create an image of a known good stick and dd it on the others. I had sticks (but mainly as cards) that managed to fail even then...

milan, to random
@milan@social.tchncs.de avatar

hmm the new machine has decreased performance, mostly noticable in php apps and when trying to jump into a container ... current theory is that it is because i chose zstd compression vs lz4 :/
for the most part mastodon seems fine tho

brot,
@brot@social.tchncs.de avatar

@milan is the zstd compression level adjustable? zstd should be comparable to lz4 when running at lower levels.

brot,
@brot@social.tchncs.de avatar

@milan oof, bugreports concerning filesystems are always scary. At least those are "only" performance issues - although those can also bring down the complete infrastructure if severe enough

milan, to random
@milan@social.tchncs.de avatar

currently running a quick cleanup through the database to free some space in the hope to speed up a potential migration to a more powerful server (in case i decide to do so) – you may experience some slowdowns depending on the table that is currently being worked on (and the base load from the fediverse is already quite high). sorry for that :tchncs:

brot,
@brot@social.tchncs.de avatar

@milan with that much ram i would recommend to try zram. Depending on the workload, it can be very much worth it. The 3 main servers at work have 256GB of RAM, and have 4x 20GB zram swap active. There is 22,5GB worth of swap in them, which only occupies 11GB. vm.swappiness is set to 30. This lead to lower iowait while having memory pressure and glusterfs doing its thing.

milan, to random
@milan@social.tchncs.de avatar

annoying how there are times where i think the hw for this instance is enough or even slightly too much, then idk musk or maybe even spez say something and i am considering a server upgrade and wonder if what i have in the basket would be sufficient

brot,
@brot@social.tchncs.de avatar

@milan can't tease us like that, what are the current specs and whats waiting in the basket?

brot,
@brot@social.tchncs.de avatar

@milan I am running a 5 year old ex40, back when nvme was not available for such small servers. I am however still waiting for the recent drops in nvme costs to trickle down to hetzner as I would like to have just everything on SSD (including terrabytes of nextcloud stuff)

milan, (edited ) to random
@milan@social.tchncs.de avatar

I think it's still going to be a while until it is relevant for our server(s) in particular, but if the new Meta project wanted to add us to their federation whitelist or just simply would open up its federation, would you as a user accept this to happen, assuming they moderate properly and the instance won't die due to load mid-term?
:tchncs:

brot,
@brot@social.tchncs.de avatar

@milan imo there are two aspects in that case: 1.I always encouraged everyone to use open protocols and oss to communicate. Never have all of my contacts been on the "cool" protocol (IRC, xmpp, signal, matrix) nor were those who did agree to get something "not WhatsApp" happy to change to the new cool protocol. Since matrix it has been easier to reach my contacts and I love that. Meta sure has the potential to add some more people on an open protocol. I have never been on any of their services.

milan, to random
@milan@social.tchncs.de avatar

there is a chance, will bring me peace ... once i can selfhost the backupserver :) – some rought edges but overall pretty smooth so far

brot,
@brot@social.tchncs.de avatar

@milan bitte weiterhin berichten. Ich hab gerade viel um die Ohren und kann mir das nicht ansehen, hab aber Interesse an dem Tool :)

  • All
  • Subscribed
  • Moderated
  • Favorites
  • random
  • uselessserver093
  • Food
  • aaaaaaacccccccce
  • test
  • CafeMeta
  • testmag
  • MUD
  • RhythmGameZone
  • RSS
  • dabs
  • KamenRider
  • KbinCafe
  • Ask_kbincafe
  • TheResearchGuardian
  • Socialism
  • oklahoma
  • SuperSentai
  • feritale
  • All magazines