Yeah, now that you mention it, I recall figuring that out also at some point. I think I did it originally when I was debugging the problem without the container. ________________________________ From: Eugen Block <eblock@nde.ag> Sent: Friday, May 3, 2024 8:37 AM To: Wyll Ingersoll <wyllys.ingersoll@keepertech.com> Cc: ceph-users@ceph.io <ceph-users@ceph.io> Subject: Re: [ceph-users] cephadm custom crush location hooks Hm, I wonder why the symlink is required, the OSDs map / to /rootfs anyway (excerpt of unit.run file): -v /:/rootfs So I removed the symlink and just added /rootfs to the crush location hook: ceph config set osd.0 crush_location_hook /rootfs/usr/local/bin/custom_crush_location After OSD restart the OSD finds its correct location. So I actually only need to update the location path, nothing else, it seems. Zitat von Eugen Block <eblock@nde.ag>:
I found your (open) tracker issue:
https://tracker.ceph.com/issues/53562
Your workaround works great, I tried it in a test cluster successfully. I will adopt it to our production cluster as well.
Thanks! Eugen
Zitat von Eugen Block <eblock@nde.ag>:
Thank you very much for the quick response! I will take a look first thing tomorrow and try that in a test cluster. But I agree, it would be helpful to have a way with cephadm to apply these hooks without these workarounds. I'll check if there's a tracker issue for that, and create one if necessary.
Thanks! Eugen
Zitat von Wyll Ingersoll <wyllys.ingersoll@keepertech.com>:
I've found the crush location hook script code to be problematic in the containerized/cephadm world.
Our workaround is to place the script in a common place on each OSD node, such as /etc/crush/crushhook.sh, and then make a link from /rootfs -> /, and set the configuration value so that the path to the hook script starts with /rootfs. The container that the OSDs run in has access to /rootfs and this hack allows them to all view the crush script without having to manually modify unit files.
For example:
1. put crushhook script on the host OS in /etc/crush/crushhook.sh 2. make a link on the host os: $ cd /; sudo ln -s / /rootfs 3. ceph config set osd crush_location_hook /rootfs/etc/crush/crushhook.sh
The containers see "/rootfs" and will then be able to access your script. Be aware though that if your script requires any sort of elevated access, it may fail because the hook runs as ceph:ceph in a minimal container so not all functions are available. I had to add lots of debug output and logging in mine (it's rather complicated) to figure out what was going on when it was running.
I would love to see the "crush_location_hook" script be something that can be stored in the config entirely instead of as a link, similar to how the ssl certificates for RGW or the dashboard are stored (ceph config-key set ...). The current situation is not ideal.
________________________________ From: Eugen Block <eblock@nde.ag> Sent: Thursday, May 2, 2024 10:23 AM To: ceph-users@ceph.io <ceph-users@ceph.io> Subject: [ceph-users] cephadm custom crush location hooks
Hi,
we've been using custom crush location hooks for some OSDs [1] for years. Since we moved to cephadm, we always have to manually edit the unit.run file of those OSDs because the path to the script is not mapped into the containers. I don't want to define custom location hooks for all OSDs globally in the OSD spec, even if those are limited to two hosts only in our case. But I'm not aware of a method to target only specific OSDs to have some files mapped into the container [2]. Is my assumption correct that we'll have to live with the manual intervention until we reorganize our osd tree? Or did I miss something?
Thanks! Eugen
[1] https://docs.ceph.com/en/latest/rados/operations/crush-map/#custom-location-... [2] https://docs.ceph.com/en/latest/cephadm/services/#mounting-files-with-extra-... _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io