[
https://issues.apache.org/jira/browse/HBASE-29161?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17933010#comment-17933010
]
Duo Zhang commented on HBASE-29161:
-----------------------------------
Good news that I've finished one round of rolling upgrading test from 2.6.2 to
3.0.0-beta-2-SNAPSHOT.
There were 2 masters and 5 region servers, and I run a LTT job while rolling
upgrading.
{noformat}
./bin/hbase ltt -write 4:1024 -num_keys 10000000 -compression SNAPPY -tn
test_ns:test_tn -region_replication 3 -num_regions_per_server 20
{noformat}
It created 100 regions each with 3 replicas, so there were 300 regions on the
cluster.
Then I upgraded 2 masters, and then 5 regionservers, one by one. For masters, I
just killed the old one and started with the new version. For regionservers, I
moved regions out and then restarted the region server with new version and
finally moved regions back, with the RegionMover class. There were some
failures while moving regions back but anyway, it did not affect the upgrading
process and correctness.
After fully upgraded to 3.0.0-beta-2-SNAPSHOT, I used LTT to verify the result.
{noformat}
./bin/hbase ltt -read 100 -num_keys 10000000 -tn test_ns:test_tn
-region_replication 3
./bin/hbase ltt -read 100 -num_keys 10000000 -tn test_ns:test_tn
-region_replication 3 -region_replica_id 1
./bin/hbase ltt -read 100 -num_keys 10000000 -tn test_ns:test_tn
-region_replication 3 -region_replica_id 2
{noformat}
There were no errors, everything worked fine.
This is a very good start. Next I will try to test rolling upgrading while
memstore replication is enabled, and also while replication between clusters
are enabled.
Thanks.
> Test rolling upgradeing from 2.x
> --------------------------------
>
> Key: HBASE-29161
> URL: https://issues.apache.org/jira/browse/HBASE-29161
> Project: HBase
> Issue Type: Sub-task
> Components: integration tests
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
>
> At least we should test rolling upgrading from 2.6.x, with some changed
> features like region replication and replication enabled, to see if the
> migration works as expected.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
