Project

General

Profile

action #71575

[osd][alert] limited /assets - idea: ask EngInfra for slow+cheap storage from central server for /assets/fixed only

Added by okurz 4 months ago. Updated 3 months ago.

Status:
Resolved
Priority:
Normal
Assignee:
mkittler
Target version:
openQA Project - Ready
Start date:
2020-09-02
Due date:
% Done:

0%

Estimated time:

Description

Motivation

/assets on osd is recently quite full. E.g. see #70885 . There are some ideas how to use storage more efficiently and also use our own "storage server". What might be a good intermediate solution (in parallel) would be to ask EngInfra for slow+cheap storage from central server for /assets/fixed only as the fixed assets are likely used by less jobs and we can live with a bit less performance.

Suggestions

  • Create ticket to ask EngInfra for slow+cheap storage, e.g. on netapp next to existing fast+expensive storage that we currently use (AFAIK)
  • Use new storage for specific mount point on /var/lib/openqa/share/factory/
  • Based on
/assets> find . -maxdepth 3 -type d -name 'fixed'
./factory/hdd/fixed
./factory/iso/fixed
./factory/repo/fixed

use the new storage for each of the above

Related issues

Related to openQA Infrastructure - action #69667: missing monitoring data for vde after partitions where reorderedResolved2020-08-06

Copied from openQA Infrastructure - action #70885: [osd][alert] flaky file system alert: /assetsWorkable2020-09-02

Copied to openQA Infrastructure - action #73165: [osd] Consolidate "expensive+fast" and "cheap+slow" storage after realizing vdc is "cheap+slow" as wellResolved2020-09-02

History

#1 Updated by okurz 4 months ago

  • Copied from action #70885: [osd][alert] flaky file system alert: /assets added

#2 Updated by okurz 4 months ago

  • Assignee set to mkittler

mkittler as discussed, please create an EngInfra ticket in the normal "infra" queue, put osd-admins@suse.de in CC, put "[openqa]" in the subject line, ask for slow+cheap (rotating disk) storage to our osd vm as additional partition or device. I would suggest something like at least 2TB if that is feasible. More is also useful.

#4 Updated by mkittler 4 months ago

The new disk is there (/dev/vde). It has 2 TiB. I can copy over fixed assets of the 3 assets types and bind-mount the directories.

#5 Updated by okurz 4 months ago

  • Related to action #69667: missing monitoring data for vde after partitions where reordered added

#6 Updated by mkittler 4 months ago

  • Status changed from Workable to Feedback
  • Formatted /dev/vde as xfs in accordance with the regular assets partition and mounted as /space-slow.
  • Copied fixed isos over: rsync -aHP /assets/factory/iso/fixed/ /space-slow/assets/factory/iso/fixed/
  • Made use of the fixed asset at their new location¹.
  • Updated /etc/fstab.

It seems to work. If no problems come up we can delete /assets/bak and move the other fixed asset directories similarly.

¹

martchus@openqa:~> cat /home/martchus/mount-assets.sh 
#!/bin/sh
set -ex
rsync -aHP /assets/factory/iso/fixed/ /space-slow/assets/factory/iso/fixed/
systemctl stop openqa-webui openqa-gru
rsync -aHP /assets/factory/iso/fixed/ /space-slow/assets/factory/iso/fixed/
mv /assets/factory/iso/fixed /assets/bak
mkdir /assets/factory/iso/fixed
mount --bind /space-slow/assets/factory/iso/fixed /assets/factory/iso/fixed
systemctl start openqa-webui openqa-gru

#7 Updated by mkittler 4 months ago

  • Status changed from Feedback to In Progress

I've removed /assets/bak and I'm continuing moving fixed HDDs and repos to the new storage.

#8 Updated by okurz 3 months ago

I will extend the quota for SLE15 on osd a bit and inform sebchlad et al: "We moved some "fixed" assets to a separate storage sink so we have a bit more space to spare for assets. I increased the size limit for assets of the SLE15 parent job group on openqa.suse.de from 980GB to 1200GB. Still, any optimization on the side of HA images would be appreciated as well as the others ideas we suggested". Also SAP+HA team has created https://infra.nue.suse.com/SelfService/Display.html?id=178140 to ask for +2TB for /assets. I commented in the ticket that as alternative cheap+slow can be increased in size partially.

#9 Updated by mkittler 3 months ago

I moved fixed HDDs as well and will remove /assets/bak (currently still holding 1.4 TB fixed HDDs) soon. For now I'll keep the repos where they are because /assets is at 60 % (when subtracting /assets/bak) and the new slow storage already at 79 %.

#10 Updated by mkittler 3 months ago

I've been deleting the backups. With

/dev/vdc        5,0T    3,2T  1,9T   63% /assets
/dev/vdd        5,0T    3,7T  1,4T   73% /results
/dev/vde        2,0T    1,6T  441G   79% /space-slow

we have the partitions kind of balanced out so I'm leaving it like this. I'll extend our file systems alert to cover /space-slow.

#12 Updated by mkittler 3 months ago

  • Status changed from In Progress to Resolved

Merged identical SR from okurz for the alert.

#13 Updated by okurz 3 months ago

  • Copied to action #73165: [osd] Consolidate "expensive+fast" and "cheap+slow" storage after realizing vdc is "cheap+slow" as well added

Also available in: Atom PDF