Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Revert "Switch to etcd snap to 3.2" #118

Merged
merged 1 commit into from
Apr 6, 2018
Merged

Conversation

Cynerva
Copy link
Contributor

@Cynerva Cynerva commented Apr 6, 2018

Reverts #115

Occasionally seeing etcd broken after a charm upgrade to edge:

etcd/0*                   active    idle   1        18.233.158.232  2379/tcp  Errored with 0 known peers
etcd/1                    active    idle   2        54.242.230.117  2379/tcp  Errored with 0 known peers
etcd/2                    active    idle   3        52.54.208.170   2379/tcp  Errored with 0 known peers

This is happening because etcd is automatically and incorrectly upgraded to v3.2.

etcd is in a crash loop, logging this:

panic: recovering backend from snapshot error: database snapshot file path error: snap: snapshot file doesn't exist

This doc indicates that you should upgrade to 3.0 first, and not upgrade to 3.2 until you have v3 data: https://github.com/coreos/etcd/blob/master/Documentation/upgrades/upgrade_3_0.md

NOTE: When migrating from v2 with no v3 data, etcd server v3.2+ panics when etcd restores from existing snapshots but no v3 ETCD_DATA_DIR/member/snap/db file. This happens when the server had migrated from v2 with no previous v3 data. This also prevents accidental v3 data loss (e.g. db file might have been moved). etcd requires that post v3 migration can only happen with v3 data. Do not upgrade to newer v3 versions until v3.0 server contains v3 data.

This issue indicates the same: etcd-io/etcd#9480

We need to revert back to 2.3 as the default so that charm upgrades don't result in a broken cluster. We will have to come up with a new way to make 3.2 the default on new deployments without breaking upgrades.

@kwmonroe kwmonroe merged commit a54b6a2 into master Apr 6, 2018
@kwmonroe kwmonroe deleted the revert-115-feature/etcd32 branch April 6, 2018 19:26
@ktsakalozos
Copy link
Member

I am afraid it is not only the upgrade path to 3.2 that is unstable.

We have incidents where non-upgraded 3.2 releases are misbehaving: https://github.com/juju-solutions/bundle-canonical-kubernetes/issues/541 (also builds in our CI show some flaky behaviour)

Should we also revert canonical/etcd-snaps@9e52ed5 @wwwtyro ?

@wwwtyro
Copy link
Contributor

wwwtyro commented Apr 10, 2018

I'm not sure I see why it'd be causing a problem with new installations, but yeah, we should remove the upgrade from everything after 3.1, so we should revert these:

I'll make the reversion PRs.

@wwwtyro
Copy link
Contributor

wwwtyro commented Apr 10, 2018

@Cynerva pointed out we should also revert the 3.1 change, so I added that one as well.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
4 participants