1 mode in cluster 1.8.0r, 1 bucket
upgrade on 2.0.2-769
server is clean and not initialized
I see the same on Linux
Upgrade 1.8.*-> 2.0.2-769 failed
looks like it's only for offline upgrade
Verify it is working for ubuntu 64.
1. first install 1.8.1 release.
2. upgrade to 2.0.2_778
Verified that beam process doesn't start.
3. sudo /etc/init.d/couchbase-server start. Verified that beam starts and server works fine.
for server-overlay-deb/etc/couchbase_init.d.tmpl, compare it between 2.0 and 2.0.2, there is bug MB-7915. Not sure if it causes the problem.
Bin, is this fixed? can you change to 'resolved-fixed' if it's ready for QE? Thanks.
No, it is not fixed yet. We haven't figured out the root cause yet.
reproduced against 2.0.2-779
all tests from MultiNodesUpgradeTests classe failed with error 'ServerUnavailableException: unable to reach the host @ IP' after upgrade
related to MB-8041. this is 64-bit, MB-8041 is 32-bit.
per Bin, priority is not a blocker – can be downgraded since there is no data loss.
per PM, this still needs to be fixed in 2.0.2 if there's no workaround.
per bug scrub, making this a blocker now.
As far as I can see regression happened in build 2.0.2-758:
One of ns_server changes is critical and quite possible that it affected upgrade path (not necessarily directly):
Apart from massive changes in ns_server there was a couple of patches related to CentOS package:
From my perspective they seem to be harmless.
Also I observe the issue in both online and offline upgrade cases.
I'd love to get any feedback from ns_server team regarding this issue.
Per discussion Aliaksey will take it and I'll take perf regression.