<02UltimateRiff> I was just about to ask the same thing, relaxoMob~
bitmap
relaxoMob, UltimateRiff: hopefully it can be deployed to beta tomorrow, and then have an extended testing period, probably being released to production on June 3rd
hope so, just need to submit a couple more PRs (one to make sure the new edit types don't ISE if you try to load them on the main site, and another to update some configuration)
ericd23_: Yes, good reminder, I will add them in a bit.
lucifer
yvanzo: hi! i took care of the alerts on datasethoster but we haven't really updated brainzbot-core in a year or so because it is very outdated and the original repo we forked it is not maintained anymore. for the short term, i think we should leave it as is and in the longer term, we will probably have to rewrite our own bot from scratch.
zas, atj: hi! when would you be available to discuss about setting up our matrix homeserver?
mayhem
awww yisss, please.
I think we settled on getting two VMs to get the job done.
lucifer
yup. need to go over the details of the setup once, setting up backups and possibly co-ordinating with libera.chat admins for extended rate limits so on.
zas
lucifer: now?
lucifer
sure
zas
did you already define resources needed in terms of disk space/cpu/ram?
yvanzo
hi lucifer, fair enough, thanks for having looked at it :)
lucifer
zas: looking at some of the cloud offerings for hosting matrix homeservers, i am thinking 8GB ram and 250GB storage.
we intend to use dedicated hetzner vms right?
CPX31, CCX23 hetzner VMs look good to me.
mayhem
do you know if the matrix is CPU intensive?
because I see the dedicated instances as not a good deal.
well, they weren't a good deal, but hetzner no longer has a 30eur/month bare metal machines. :(
two CPX31s do seem like a pretty good idea.
how about CAX21 or CAX31? I hear good things about the arm64 server. but I am not sure that everything we need to install is available for arm64
lucifer
reading from other people's experiences online, it doesn't seem to be cpu intensive.
mayhem
great. then lets not get a dedicated instance, shouldn't be needed.
lucifer
yeah i saw the arm ones. but i am not sure everything supports it at the moment.
mayhem
how difficult would it be to check that>
?
I could get you a test machine and if its a pain, we dump it and move to amd
zas: did we ever figure out if we still need musicbrainz-docker-test ?
lucifer
zas: should be okay i think. as for redundancy, let me check.
mayhem
been created 3 years ago.
lucifer: 65.21.242.117
log in as root.
lucifer
thanks!
mayhem
zas: do you know if our metabrainz server setup routine can run on arm64?
atj
what do you mean by metabrainz server setup routine?
ansible bootstrap or something else?
lucifer: there are arm64 docker images AFAICS
mayhem
atj: not familiar with all the ansible bits... but the process that takes a bare machine and makes it a metabrainz machine with all the accounts, firewall, etc.
<lucifer> "zas, atj: it seems redundancy is..." <- I have my synapse in a VM and the whole VM in HA-setup. redundancy would mean more workers/reverse proxies or those in a combination of these synapse workers. but downtime isn't that big issue, every federation caches and pulls the delta when it comes back online.
maybe dendrite could be worth a shot, dunno, never tried it
lucifer
mr44er[m]: i see, do the VMs share a database?
atj
I use dendrite for a personal homeserver and it's missing a lot of features
mr44er[m]
<lucifer> "mr44er: i see, do the VMs..." <- the db is exclusive for synapse within the VM, but it would not be necessary
I was just lazy and like everything in the VM
lucifer
i think to be fully HA we would need a db on each VM too. so iiuc the way your setup is running atm.
mr44er[m]
if I remember correctly ~10 days of downtime was default to keep/pull messages back
lucifer
how do the synapse instances on multiple VMs sync with each other?
mr44er[m]
like every federation/matrix server
but true HA from within synapse isn't possible afaik.
but multiple postgresql-servers should be possible. if you point synapse to front/mainserver it should work.
lucifer
a bit confused right now. i am looking for HA outside of synapse.
to be clear.
but the two options i found were HA within synapse which seems hard to get.
mr44er[m]
I have one instance, one VM with synapse and postegresql.
lucifer
makes sense. how do you get HA with it?
mr44er[m]
and this (whole) VM configured as HA, independently from synapse etc
proxmox cluster
lucifer
ah i see, never heard the name before but is it like vm snapshotting in almost real time?
mr44er[m]
one server goes down, no problem. VM switches server, no downtime, no ping lost
yes
lucifer
got it, makes sense.
mr44er[m]
true HA, no downtime, I recommend ceph 😉
lucifer
hehe. i think too much complexity for our use case.
i was hoping for something like a postgres master/standby system where we could flip a switch to failover manually or maybe automatically.
mr44er[m]
yeah, this would be the best solution. a combination of this I use with opnsense in addition to the HA. updates without downtime, because instance1 or instance2 always overtakes automatically
with postgres it would work, but synapse is only single instance right now
lucifer
right i meant if synapse could function like that. but that doesn't seem possible.
the options you have mentioned look good too but i personally have no idea if its worth the additional maintenance effort needed. probably the best for zas or atj to decide.
atj
lucifer: no, just use a single instance and be done with it
lucifer
cool
atj
if this were critical functionality then it might be worth investing the effort to achieve HA
lucifer
makes sense
thanks for help though mr44er[m]! :)
atj
if you want me to bootstrap a server let me know
mr44er[m]
<lucifer> "the options you have mentioned..." <- it is my workaround for the lack of synapse's feature regarding redundancy/HA whatsoever 🙂