This is an automated email from the ASF dual-hosted git repository.
alexey pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/kudu.git
The following commit(s) were added to refs/heads/master by this push:
new fd8af5e94 [docs] Elaborate rack-aware rebalancing
fd8af5e94 is described below
commit fd8af5e9422453407a5b0e6a5951ab2452eb20da
Author: Abhishek Chennaka <[email protected]>
AuthorDate: Tue Jan 7 22:32:09 2025 -0800
[docs] Elaborate rack-aware rebalancing
Adding some examples and a small note based on field feedback on the
documentation of running rebalancer in a rack-aware cluster.
Change-Id: I6724c8cdd69167fabf51b66a462dfa25338c057d
Reviewed-on: http://gerrit.cloudera.org:8080/22312
Tested-by: Kudu Jenkins
Reviewed-by: Alexey Serbin <[email protected]>
---
docs/administration.adoc | 95 ++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 95 insertions(+)
diff --git a/docs/administration.adoc b/docs/administration.adoc
index 3ec10d0bd..ec939b02e 100644
--- a/docs/administration.adoc
+++ b/docs/administration.adoc
@@ -1584,10 +1584,105 @@ tool breaks its work into three phases:
location, as if the location were a cluster on its own. Use the
`--disable_intra_location_rebalancing` flag to skip this phase.
+Note: Each of the above rebalancing phases can be independently skipped using
the
+corresponding flags.
+
By using the `--report_only` flag, it's also possible to check if all tablets
in
the cluster conform to the placement policy without attempting any replica
movement.
+[[rebalancer_with_rack_awareness_example]]
+==== Examples
+The behavior of each of the these flags can be better explained with examples.
+Consider a case where there are 3 locations and 3 tablets (9 replicas
+total) and running the tool with the below flags (effectively running only the
placement policy
+fixer):
+
+ --disable_cross_location_rebalancing --disable_intra_location_rebalancing
+
+[[rebalancer_example_policy_fixer]]
+.Before running the tool
+[options="header"]
+|===
+| Location A | Location B | Location C
+| Replica X | Replica Y | Replica Z
+| Replica X | Replica Y | Replica Z
+| Replica X | Replica Y | Replica Z
+|===
+
+[[rebalancer_example]]
+.After running the tool with the flags
+[options="header"]
+|===
+| Location A | Location B | Location C
+| Replica X | Replica X | Replica X
+| Replica Y | Replica Y | Replica Y
+| Replica Z | Replica Z | Replica Z
+|===
+
+Notice the replicas of every tablet are now spread across all the 3 locations.
+
+Next, let's consider the below tablet distribution and running the tool with
the below flags
+(effectively running only the cross-location rebalancing):
+
+ --disable_policy_fixer --disable_intra_location_rebalancing
+
+[[rebalancer_example_cross_location_rebalancing]]
+.Before running the tool
+[options="header"]
+|===
+| Location | Number of replicas across all tables in the location
+| A (5 tablet servers) | 15
+| B (5 tablet servers) | 18
+| C (5 tablet servers) | 21
+|===
+
+[[rebalancer_example]]
+.After running the tool with the flags
+[options="header"]
+|===
+| Location | Number of replicas across all tables in the location
+| A (5 tablet servers) | 21 replicas
+| B (5 tablet servers) | 21 replicas
+| C (5 tablet servers) | 21 replicas
+|===
+
+Notice the number of replicas in each of the locations is now equal.
+
+Continuing the above example, let's examine location A before and after
running the
+tool with the flags (effectively running only the intra-location rebalancing):
+
+ --disable_policy_fixer --disable_cross_location_rebalancing
+
+[[rebalancer_example_intra_location_rebalancing]]
+.Before running the tool
+[options="header"]
+|===
+| Tablet server | Number of replicas across all tables in the server
+| TS_1 | 3
+| TS_2 | 5
+| TS_3 | 8
+| TS_4 | 4
+| TS_5 | 1
+|===
+
+[[rebalancer_example]]
+.After running the tool with the flags
+[options="header"]
+|===
+| Tablet server | Number of replicas across all tables in the server
+| TS_1 | 4
+| TS_2 | 5
+| TS_3 | 4
+| TS_4 | 4
+| TS_5 | 4
+|===
+
+Notice the number of replicas in each tablet server is now balanced.
+
+Note: In the above example all the three flags can be enabled together. We
individually enabled
+them in separate runs to aid understanding.
+
[[tablet_server_decommissioning]]
=== Decommissioning or Permanently Removing a Tablet Server From a Cluster
