ClusterLabs / resource-agents

Combined repository of OCF agents from the RHCS and Linux-HA projects
GNU General Public License v2.0
493 stars 581 forks source link

Resource agent "mdraid" need "md_dev=auto" setting. #1800

Open mlaguilharre opened 2 years ago

mlaguilharre commented 2 years ago

We have thousands resource agent Raid1 that contains each several soft raid device. They are managed thanks to the parameters raidconf="/etc/mdadm.conf" raiddev=auto. The /etc/mdamd.conf contains several entries for each soft raid device:

eg:

# /etc/mdadm.conf
ARRAY /dev/md/test-system metadata=1.2 name=node1:test-system UUID=ba616f8a:78da4bee:618dbdf6:cf9889f1
ARRAY /dev/md/test-data1 metadata=1.2 name=node1:test-data001 UUID=b805c2c2:b88442c3:b23bfd5f:b052b45b
ARRAY /dev/md/test-data2 metadata=1.2 name=:node1:test-data2 UUID=13fc4c7a:97067415:85dca6fe:20cc2f8

With latest pacemaker version, "Raid1" Resource Agent is obsoleted by "mdraid", unfortunately there is no equivalent parameter in "mdraid" for raiddev=auto.

If we need to migrate to mdraid, it needs to create a Resource Agent for each Soft Raid device which is cumbersome.

zhaohem commented 2 years ago

In my view, the 'auto' function should be supported in mdraid RA.

oalbrigt commented 2 years ago

I think that's a per distro decision. Pacemaker should support both Raid1 and mdadm.

I've asked the dev of the agent whether we intend to add this feature to mdadm.

oalbrigt commented 2 years ago

There's no plans to add this feature to the mdraid agent.

zhaohem commented 2 years ago

Thanks for the info. let's keep opening this issue. If I have time, I will submit a patch. I estimate the time is some months later.