[
https://issues.apache.org/jira/browse/CASSANDRA-13701?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17177965#comment-17177965
]
Alexander Dejanovski commented on CASSANDRA-13701:
--------------------------------------------------
I've identified several issues today:
* ccm uses a hardcoded 30s timeout when waiting for events (like nodes to
start) which doesn't work with the additional wait times that come with the new
token allocation algorithm. Fix is
[here|https://github.com/riptano/ccm/commit/8a91a5aa49473211863a1fb7a980206e5222ce5d].
* ccm starts all nodes at the same time when cluster.start() is invoked, which
creates clashes when the new token allocation algorithm is used and makes some
tests flaky. Starting them sequentially using [this
fix|https://github.com/riptano/ccm/commit/e6e4abcff375debde8195104c5cffd1cecb8d6cf],
allowed all the bootstrap dtests to pass.
* [~jeromatron]'s branch is missing some commits in the current trunk that fix
other failing dtests. Rebasing it over trunk is necessary to get them all to
pass
* Adding a few seconds of sleep in
[bootstrap_test.py::TestBootstrap::test_simultaneous_bootstrap|https://github.com/adejanovski/cassandra-dtest/blob/master/bootstrap_test.py#L769-L771]
allows the test to pass.
I'm currently rerunning all dtests with the various fixes to see if I still get
failures. I'll follow up on monday and hopefully push PRs to ccm and
cassandra-dtests that will allow the patch to be applied (there are conflicts
though so a rebase will be necessary).
A follow up discussion and ticket will probably be necessary because the new
token allocation algorithm and concurrent bootstraps aren't working nicely
together.
> Lower default num_tokens
> ------------------------
>
> Key: CASSANDRA-13701
> URL: https://issues.apache.org/jira/browse/CASSANDRA-13701
> Project: Cassandra
> Issue Type: Improvement
> Components: Local/Config
> Reporter: Chris Lohfink
> Assignee: Alexander Dejanovski
> Priority: Low
> Fix For: 4.0-alpha
>
>
> For reasons highlighted in CASSANDRA-7032, the high number of vnodes is not
> necessary. It is very expensive for operations processes and scanning. Its
> come up a lot and its pretty standard and known now to always reduce the
> num_tokens within the community. We should just lower the defaults.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
