monkey ansh ooooh new graph!! Really cool. Love the colours
P.S. I wonder if we should be careful of not dropping the artist country graph down too far? We’ve had a few people shout that out as particular useful/different to other services
leftmostcat (UTC-8): hiya! Are you the author of the perl module unidecode?
\- has quit
lucifer[m]
zas: hi! i think the spark servers upgrade didn't go right? marlon shows only 450G of total space
all servers i checked actually.
zas[m]
that's the storage they have (2 SSDs of ~500gb in RAID1)
lucifer[m]
i remember it used to 1TB before the server upgrade
on each server
mayhem[m] suspects RAID-0 vs RAID-1 issue
mayhem[m]
they were striped, not mirrored, right?
lucifer[m]
i see yeah that's possible. i think we can do RAID-0 for spark servers because HDFS does provide redundancy for the data in the spark cluster.
zas[m]
oh, that would explain this
lucifer[m]
that said there have been disk issues with the cluster in the past.
but 1 TB of storage is needed on almost all nodes to avoid running into disk issues. so we'd have to add new disks if we want to keep RAID-1.
zas[m]
ok, I'll see if we can convert without reinstalling, perhaps tricky, but I can try
I can perhaps do it from rescue, can I stop marlon in order to test?
lucifer: ?^^
lucifer[m]
zas: its currently running a job but i don't think it will succeed with the current diskspace anyway. so yup feel free to restart any/all of the spark servers.
[@zas666:matrix.org](https://matrix.to/#/@zas666:matrix.org) another question, we need about 8 VMs for spark processing of MLHD dumps for probably about a week
Whwn do you think we can do that?
zas[m]
If mayhem is ok (money-wise), we can. Which VM specs do you need?