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

Update pipe_verify_handoff_blocking to handle vnode shutdown #1093

Open
wants to merge 144 commits into
base: riak/2.0
Choose a base branch
from

Conversation

JeetKunDoug
Copy link
Contributor

Occasionally the test would fail with an error where a vnode was transferred after the vnode proxy had gotten its pid but before the work could be sent.

lordnull and others added 30 commits December 30, 2014 16:37
rpc or other processes will cause one or two persistent processes to start up during
the test, but the number shoould not increase after that. Altered the test to check for
massive increase in processes over a larger number of faults rather than an equality
of process_count over a relatively small number of faults.
* Backed out the change to riak_core_handoff_sender_intercepts which had
  a modified intercept
* Modified the handle_command intercept to ensure we don't count the
  vnode crashes that were happening before the fix
* Reverted the test to the original, but check to make sure we can read
  all the entries, including the ones we added during handoff
update some yokozuna tests from work done in riak/2.0 branch for 2.0.…
(by not waiting for handoffs, just pending changes, as the old test did)
* add new types for dialyzer
* add test to reload schema and overwrite it on type change
* test nested search (e.g. nested json objects)
* test re-putting fields, include crdt map workaround, discussed in
https://gist.github.com/sdebnath/36c235e042cb35db7d1f
* increase aae concurrency for core_prop test
The verify_snmp_repl test reproduces the error condition (one cluster
configured for realtime replication with two or more other clusters) and
passes if and only if the riak_snmp_stat_poller module correctly handles
this configuration.

https://bashoeng.atlassian.net/browse/RIAK-1884
basho/riak_snmp#27
…ugh handle_command, to ensure we are testing the case statements in riak_kv_vnode:handle_command with write_once puts.
Add basic cluster converge test to try to capture issues with cluster convergence outside of other tests.
@nickelization
Copy link
Contributor

@JeetKunDoug There's a comment around line 110 that seems to refer to the old behavior involving vnode_down. Seems like it should be deleted.

@nickelization
Copy link
Contributor

Gotta run out for a bit, but other than that one comment, it all looks good to me. Once that's fixed, you can consider this +1'd.

@JeetKunDoug
Copy link
Contributor Author

JeetKunDoug commented Jun 16, 2016

After further code review, realized that the race condition mentioned isn't really what was going on - fixed the bug in riak_pipe:queue_work_wait in basho/riak_pipe#106 (RIAK-1723) (RIAK-1830). This test should fail without that fix, and pass with it.

fadushin and others added 23 commits July 7, 2016 15:47
# Conflicts:
#	tests/yz_core_properties_create_unload.erl
- update crdt cap and converge tests for hlls
- temp rebar updates for riak-client; *will update w/ correct version*
upon mergies
Update verify_aae test to verify throttle config
This can be disabled by setting the `giddyup_post_debug_logs` config
value to `false` (it is `true` by default).
This pulls in the latest client code, as well as newer riak_pb stuff
which is needed for the latest develop-2.2 Riak code.
Update client libraries to use develop-2.2
All the HLL stuff has been merged into the client develop-2.2 branches,
so we can go ahead and update these dependencies to match that.
Somehow managed to pass on 2.0.7, but shouldn't have (and doesn't in 2.2.0).
Add back call to stop nodes before deleting index directories.
Wait for CRDT search results, rather than failing after the first attempt.
Use intercepts to provoke an otherwise-occasional error where a vnode 
has been transferred after the vnode proxy has gotten its pid
but before the work can be sent. Will fail until fix to
`riak_pipe_vnode:queue_work_wait` is merged. See basho/riak_pipe#106
for the fix.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet