Lowering max_retries from 5 to 2 to catch build errors faster
Created by: BjoernT
RPC-openstack utilizes the install_bits function from OSA which is a wrapper around successerator and by default it uses a MAX_RETRIES parameter of 5. I'm asking rpc-openstack to fail more early in case repeating errors occur. It has been our experience if playbooks fail more than one, there's usually a issue we have to take care off. There is then no need to wait for 3 more failures before we even get a chance to fix and rerun the playbooks again.
Imported comments:
By BjoernT on 2015-10-13 19:26:06 UTC
Pull requests see https://github.com/rcbops/rpc-openstack/pull/480 and https://github.com/rcbops/rpc-openstack/pull/481
By d34dh0r53 on 2015-10-14 14:11:05 UTC
Thanks @bjoernt!
By rackertom on 2015-10-22 21:29:43 UTC
Given the comments in #480 and #481 it seems the intent was never to actually use scripts-library.sh in the deploy.sh script, and probably not in any other script in the rpc-openstack repo.
If @git-harry @bjoernt @odyssey4me agree that we should remove scripts-library.sh from the deploy.sh script, I have a patch I can easily submit for review. This would remove the use of install_bits which in turn would also remove the use of successerator. This does mean that there would be no automatic retry during a deploy.sh run, but it seems like in our case this is actually preferable.
By BjoernT on 2015-10-22 21:33:54 UTC
Yes this is alright by me.
By git-harry on 2015-10-27 10:34:26 UTC
I've closed pull requests #480 and #481. This issue is now pending a pull request from @rackertom
By prometheanfire on 2015-11-02 20:32:09 UTC
just needs backport now