#metabrainz

Posted the Picard 3.0 Q&A blog post, before another post pops up to displace it again

outsidecontext ^

Backing up SolrCloud 9 collections.

yvanzo: did it work? I see folders on the nodes?

It seems so, I’m setting up a standalone Solr instance on wolf to try loading them.

Also, good news is that the new Solr incremental backups should make dump creation much faster.

are you just syncing them from all the nodes into one folder?

no sync at this stage

just pre-making folders on all nodes beforehand

:/

for example, recording has 4 shards, each with a different leader, I just sent one query and they all did generate the appropriate backup files.

(one query to only one node)

[listenbrainz-server] 14anshg1214 opened pull request #2886 (03master…search-page): Add Entity Search https://github.com/metabrainz/listenbrainz-serv...

[23:39] <aerozol[m]> Thanks yvanzo, copied (with slight adjustment) to X

can we not call twitter x? can we be social disobedient enough to elon and his cronies to *continue* to call it twitter?

jeg bare spør

[listenbrainz-server] 14amCap1712 merged pull request #2872 (03master…jspf-cleanup): Jspf cleanup https://github.com/metabrainz/listenbrainz-serv...

I want kids 10, 20 years from now asking us "why ar eyou calling x twitter?" and thne the story of why never dies

[listenbrainz-server] release 03v-2024-05-29.0 has been published by 14github-actions[bot]: https://github.com/metabrainz/listenbrainz-serv...

I would very much hope there's no twitter/x in 5 years, much less 10 or 20 :p

I would hopw so? i go friends there and a community

right now it's already falling apart because of everything

(and the lack of new lupincontent but that's a different thing)

<05rustynova> I personally don't think Elon will remain CEO for long, but the platform wouldn't die. And either I call it twitter, or "X (formerly known as twitter)" because it's funnier that way

i think refusing to use a platform owned by a white supremacist would be a more suitable form of "social disobedience" than quibbling about names

^ this times 1000

rustynova :D

well "not using it" means jsut "I won't be able to interact with these people i want to interact with" so that's moot

well in that case don't try to police what other people do then

also they don't care about users not using it, they *do* care about their dumb brandname not being used

"can we not call twitter x?" = policing, now apparently

it was a polite bequest, it was not an affront to you, personally :P

stop taking it that way ;)

me criticizing bad takes is not "taking it personally"

but hey, I'm sure Elon will ban all the Nazis and stop retweeting Ian Miles Cheong if you call it Twitter enough

well that's what it comes off as. also who are you to decide whats a bad take or not :P

he definitly wont

I expect unless something changes, communities that currently live on Twitter will probably move elsewhere eventually

I guess the main question is what the somewhere is :)

yep, they definitely will, but the problem is that it will splinter, and that makes me unhappy so I'm a bit ornery :wry:

:)

reosarevok: seems obvious: Truth Social ;)

oh look, the summit weather forecast has arrived: https://www.theguardian.com/world/article/2024/...

will get worse next month, 50-55 even. should be better once the monsoon rains start by july.

🫠

CPU+NVME 🫠

ApeKattQuest: I feel like calling it X (which nobody knows) does more harm to Elon etc?

I can use ‘X (formerly Twitter)’ like most places do, when the word count allows 👍️

The X formerly known as Twitter

Xitter

yvanzo: i'll try and get NFS set up today

will mount the share under `/var/lib/solr/backups`

then i need to install rsync and set up the SSH key. might be best if we generate a new one?

*rrsync

atj: The new backup format can no longer be tampered with to load into Solr standalone mode. Will have to set SolrCloud mode even for mirrors.

Thanks, it will always be useful to have an NFS share in case we actually need backups.

Yes to generating a new key.

atj: Not even sure if that can be restored in another cluster.

yvanzo: that is stretching the meaning of the word "backup" if true

I mean the backup might also include the distribution on nodes.

atj: I noticed that `solr` user/group IDs vary across the nodes.

hmm, not sure how that happened

yvanzo: i will need to shutdown solr on each box to fix this

nfs3 vs nfs4 maybe?

nah, it's an error on my part

kk

i set the uid / gid to be 998 in the ansible configuration, but other packages took it

atj: Does beta.mb.o have to be switched back to Solr 7 in the meantime?

yvanzo: i can do it one by one, so no

it will be a good test of the resiliency actually

Yes, we should see some changes in shards’ leaders.

I saved the collection status locally for comparison.

yvanzo: great thanks. i will start on node 1 now

yvanzo: node 1 is back

nodes 2, 3 & 5 took the lead on its four shards

I need to nip out for a little bit now, but the process is straightforward so I'll just go through the others one by one when I get back

Thanks!

thanks for noticing, much better to fix this now!

[listenbrainz-server] 14Aerozol opened pull request #2887 (03master…Alphabetize-add-data): Alphabetise AddData.tsx https://github.com/metabrainz/listenbrainz-serv...

atj: lucifer: zas: how can we move the ipv6 support issue forward in order to complete the matrix migration?

mayhem: i think atj was looking into that but currently busy with solr stuff.

zas: do you know if we can assign 2 IPv6 /64 subnets to a VM?

mayhem: lol prediction came true too soon. https://www.ndtv.com/india-news/delhi-hits-52-3...

atj: I don't think so. Why would it be needed?

zas: need to delegate a prefix to the matrix irc bridge

atj, zas: not sure if it makes sense, but would it possible to to use a /48 or /32 subspace inside the /64 already assigned to the vm?

lucifer: i had another look at the issue i linked before, and it looks like that's possible

however the docker docs are a bit unclear "the base value for IPv6 needs a minimum prefix length of /64. This is due to an integer overflow in the Docker daemon"

a great couple of notes there "defining an excessive number of subnets causes the daemon to consume all available memory."

atj: ah sorry, i had it the other way around i think. i meant can we use a subdivision of the /64 subnet in docker and matrix?

lucifer: no i think it's me :P

i will need to enable ipv6 support in docker and test some things

sounds good, let me know if i can help.

lucifer: hmm, this makes me think IP_FREEBIND won't work inside Docker

tbh i'd rather use a container platform that has sensible networking support than battle with Docker

what do you think about running the service without docker?

maybe just run the irc bridge in a systemd-nspawn container?

i am not sure how to do that. 😅

i think using the ansible role with docker is best from a maintenance perspective

sticking one isolated component in a separate container shouldn't be an issue but i need to check out how it works

lucifer: 😬