Update git submodules
* Update plugins/replication from branch 'master'
to cf0aea56d2efad79bcb84df37edc32c01bf8afb7
- Merge branch 'stable-3.10'
* stable-3.10:
PushOne: Remove unused addRef() method
TasksStorage: Remove synchronized from methods
Place the replaying flag clearing in a finally
Demote delete errors when distributor is enabled
distributor: Reduce log level for no-op consolidations
Change-Id: I3ce1519c44ec6cbae7725527ea626e55401cc488
Release-Notes: skip
- Merge branch 'stable-3.8' into stable-3.10
* stable-3.8:
PushOne: Remove unused addRef() method
TasksStorage: Remove synchronized from methods
Place the replaying flag clearing in a finally
Demote delete errors when distributor is enabled
distributor: Reduce log level for no-op consolidations
Change-Id: I15eb65d0f379a6e562d3c86e1089cfb7d9e6752c
Release-Notes: skip
- Merge branch 'stable-3.5' into stable-3.8
* stable-3.5:
TasksStorage: Remove synchronized from methods
Place the replaying flag clearing in a finally
Demote delete errors when distributor is enabled
distributor: Reduce log level for no-op consolidations
Change-Id: Iaa6d4a59d5cb26879500f2d36852eaa30efc5d7b
Release-Notes: skip
- PushOne: Remove unused addRef() method
This is only used by tests and should have been removed when
addRefBatch() superseded it.
Change-Id: If9de1df29fd56a595fa4d4b2b4f5716264a67e11
Release-Notes: skip
- TasksStorage: Remove synchronized from methods
These aren't expected to be necessary for correctness.
Change-Id: Id23d3b385e289ba453638b76d6523016a5aabf2a
- Place the replaying flag clearing in a finally
This ensure that 'replaying' will be set to false even if an exception
occurs in pruneNoLongerPending(), otherwise if would be possible for the
distributor to get "stuck" and never be able to run again.
Release-Notes: Avoid opportunity to inadvertently block the distributor
Change-Id: I4628a574ad54ca6abd591138edc48e12a39447e7
- Demote delete errors when distributor is enabled
Since delete errors for Tasks is expected after recovery (when another
nodes starts up) when running with a multi-primary setup, demote them to
"atFine()" level when the distributor is enabled. This avoids errors in
the logs during normal multi-primary operation.
Release-Notes: Reduced log level for Task deletion errors in MP setups
Change-Id: Iccca2e1bd2b91b91e12a14c4e473c8a6df6fd4b7
- distributor: Reduce log level for no-op consolidations
It's expected that the replication distributor will generate many of
these log messages on a busy instance because it will repeatedly
re-schedule the same ref updates. Since those messages likely aren't
operationally interesting, reduce the log level to atFine when
scheduling is happening based on stored replication tasks and the ref is
already scheduled in the push task.
Change-Id: I07d071af848481fa84e1ba26fb8287c5e61d1d67
Release-Notes: Reduced log level for no-op distributor push consolidations
diff --git a/plugins/replication b/plugins/replication
index 3982574..cf0aea5 160000
--- a/plugins/replication
+++ b/plugins/replication
@@ -1 +1 @@
-Subproject commit 3982574d28f254b525b093c2ecae8caa47825910
+Subproject commit cf0aea56d2efad79bcb84df37edc32c01bf8afb7
