>> My question is: Is there a way to force the snapshot creation / sync
from the leader?
3.5.6 will automatically create a clean snapshot as part of server start up
process. So a snapshot should be available after initial upgrade and there
is no need to force a snapshot creation.
On Mon, Dec 9, 2019 at 2:11 AM Thomas Cooper wrote:
> Hi,
>
> I am working on upgrading a Kubernetes based 3.4.14 cluster to 3.5.6.
>
> I are hitting the issue where the new snapshot checks prevent the newly
> upgraded 3.5.6 cluster from starting. This usually happens in our test
> set-up as the 3.4.14 cluster is not very old and has not hit the snapCount
> limit yet. Typically only one node (presumable the leader) has a snapshot
> file but the other do not.
>
> I can disable the snapshot checks via the config value for the upgrade to
> 3.5.6, however we obviously need to re-enable the checks ASAP. The issue is
> making sure all of the ZK nodes have snapshots so that the checks can be
> re-enabled.
>
> My question is: Is there a way to force the snapshot creation / sync from
> the leader?
>
> As part of our upgrade process we need to restart the servers soon after
> the initial upgrade of the ZK container images to enable config changes. If
> the snapshot creation / sync hasn't happened by this point then obviously
> the checks will fail. If we can force a snapshot creation, after the
> initial upgrade, we can re-enable the checks as part of the upgrade
> process. If not, we will need to create some method to check if the
> snapshots have been created and periodically inspect the ZK server file
> system, re-enabling the checks once they all have snapshots.
>
> Any pointers would be appreciated.
>
> Thanks in advance,
>
> Tom Cooper
>