Conversation

Notices

1. Embed this notice
   :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:37:39 JST :suya:

   • karolat
   what the fuck? Did we lose half of our database?
   
   @karolat what did you do?
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 05:37:38 JST karolat

     in reply to
     @newt I started a vacuum analyze last night, the vacuum part removes parentless records :comfygeek: Did it really clear up that much space?
     :blobcathug: likes this.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 05:37:52 JST karolat

     in reply to
     @newt you know though, we can probably run repack now considering the true size of those tables now :blobidea:
     :blobcathug: likes this.
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:37:53 JST :suya:

     in reply to

     • karolat
     @karolat yes and no. Postgres doesn't track it anymore, but the space is still occupied on disk. We need to do VACUUM FULL to reclaim it.
   • Embed this notice
     Ennui (iwillbite@whinge.town)'s status on Friday, 10-Feb-2023 05:38:17 JST Ennui

     in reply to

     • karolat
     @karolat @newt literally nerd talk
     :blobcathug: likes this.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 05:38:18 JST karolat

     in reply to
     @newt we have enough to do them a few tables at a time. Now using it on activities is possible thanks to my work :blobcatcomfyuwu:
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:38:19 JST :suya:

     in reply to

     • karolat
     @karolat sadly, no. We don't have 180 gigs free.
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 05:42:31 JST :blobcathug:

     in reply to

     • karolat
     @newt @karolat i wana join the discussion, im curious how fast the db would be if i would compress activities and objects with lz4
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:32 JST :suya:

     in reply to

     • karolat
     @karolat also, I don't really want to prune it. Pleroma won't like it.
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:33 JST :suya:

     in reply to

     • karolat
     @karolat it won't. Again, the index seems to be the problem. I'll try to reindex my copy and see how long it takes before doing the same on the main database.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:34 JST karolat

     in reply to
     @newt we can prune and vacuum again to see if that gets activities small enough to unpack locally :blobcatcomfthink:
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:35 JST :suya:

     in reply to

     • karolat
     @karolat yup. But activities and objects were the bulk of it, really. The rest is minuscule.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:37 JST karolat

     in reply to
     @newt for the whole db though, right?
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:38 JST :suya:

     in reply to

     • karolat
     @karolat last I tried, it was 5 days on my workstation.
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:39 JST karolat

     in reply to
     @newt we could go offline just to vacuum full that table and then we’d have space for the rest. I wonder how long it would take :blobcatcomfthink:
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:40 JST :suya:

     in reply to

     • karolat
     @karolat i'll figure something out. Don't worry :comfycool:
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:41 JST karolat

     in reply to
     @newt how do we reindex? We need 2x free space for the repack which I guess we don’t really have
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:42 JST :suya:

     in reply to

     • karolat
     @karolat wait, I use a wrong query. But!
     
     Half of that size is indices. So, if we reindex the database, half of that shit will be gone!
     
     Honestly, this is like the first time I'm seeing that indices are larger than the data itself.
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 05:42:43 JST :suya:

     in reply to

     • karolat
     @karolat ah maybe. Ok, I'll try this out some time.
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 06:07:09 JST :blobcathug:

     in reply to

     • karolat
     @newt @karolat
     wouldn't be so sure about that
     https://www.postgresql.fastware.com/blog/what-is-the-new-lz4-toast-compression-in-postgresql-14
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 06:07:10 JST :suya:

     in reply to

     • karolat
     • :blobcathug:
     @Jain @karolat it won’t be any faster, that’s for sure
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 06:11:37 JST :blobcathug:

     in reply to

     • karolat
     @newt @karolat :blobcat3c: let me know... the big issue i have with it is taking the risk and actually test it out... if it goes wrong i will probably have issues for hours because the big tables are on a slower storage rn
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 06:11:38 JST :suya:

     in reply to

     • karolat
     • :blobcathug:
     @Jain @karolat I’ll make a new copy of our database to experiment on tomorrow. This will probably be on my checklist.
     :blobcathug: likes this.
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 06:16:09 JST :blobcathug:

     in reply to

     • karolat
     @newt @karolat do you mirror all requests to the second server to benchmark or experiment with things?
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 06:16:10 JST :suya:

     in reply to

     • karolat
     • :blobcathug:
     @Jain @karolat and this is why I first try things on a copy
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 06:19:14 JST :blobcathug:

     in reply to

     • karolat
     @newt @karolat so you would not know the actual load which then falls on the new env...
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 06:19:15 JST :suya:

     in reply to

     • karolat
     • :blobcathug:
     @Jain @karolat nope. I just see if things aren’t screwed up terminally. But mostly this is to see how long it would take e.g. to reindex or vacuum the database.
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 06:20:06 JST :blobcathug:

     in reply to

     • karolat
     • :blobcathug:
     @newt @karolat why dont u use repack?
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 06:23:53 JST :blobcathug:

     in reply to

     • karolat
     @newt @karolat i thought so too, had the idea while reading your text... im pretty sure thats possible somehow and if i got some time i might test things out... but on the other hand to create an environment like this one probably have to replication before mirroring and everything might get difficult to just setup such an experimental env
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 06:23:54 JST :suya:

     in reply to

     • karolat
     • :blobcathug:
     @Jain @karolat I haven’t gone that far. But mirroring all traffic seems like a good idea.
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 06:26:03 JST :blobcathug:

     in reply to

     • karolat
     @newt @karolat ah, i know that... same situation lead to the weird config i do now have... having everything on fast storage except activities and objects, they are on hetzners block storage which is way slower but its cheaper and costs me less
   • Embed this notice
     :suya: (newt@stereophonic.space)'s status on Friday, 10-Feb-2023 06:26:04 JST :suya:

     in reply to

     • karolat
     • :blobcathug:
     @Jain @karolat not enough space
   • Embed this notice
     :blobcathug: (jain@blob.cat)'s status on Friday, 10-Feb-2023 10:06:47 JST :blobcathug:

     in reply to

     • karolat
     @karolat @newt oh i have no idea, this was a spontaneous idea, i don't know if it would work at all...
   • Embed this notice
     karolat (karolat@stereophonic.space)'s status on Friday, 10-Feb-2023 10:06:49 JST karolat

     in reply to

     • :blobcathug:
     @Jain @newt how do you do this? :blobcatcomfthink: