arshilkxwork commented on code in PR #547:
URL: https://github.com/apache/hugegraph-doc/pull/547#discussion_r4446104169
##########
content/en/docs/quickstart/hugegraph/hugegraph-pd.md:
##########
@@ -168,8 +168,17 @@ install-snapshot timeout, and id allocation runs no
ReadIndex wait.
| `pd.initial-store-count` | `1` | `3` | Minimum number of active Store nodes.
Below this the cluster state becomes `Cluster_Not_Ready` and the cluster is
treated as unavailable. Set it to the number of stores you deploy. |
| `pd.initial-store-list` | `127.0.0.1:8500` | empty | Comma separated Store
gRPC addresses (`ip:port`) that are activated automatically when they register.
An entry may also carry a group id as `store_address/group_id`. |
| `pd.cluster_id` | not set | `1` | Cluster id, used to keep separate PD
clusters apart. |
+| `pd.balance-leader-enable` | `false` | `false` | Run partition leader
balancing periodically on the PD leader. When off, leaders move back only
through a manual `balanceLeaders` call. |
+| `pd.balance-leader-interval` | `300` | `300` | Seconds between periodic
leader balance checks. A check plans transfers only when this much time has
passed since the last plan, including a manual one. Must be positive. |
+| `pd.balance-leader-max-skew` | `2` | `2` | A check balances leaders only
when the most leaders on one Store minus the fewest on another exceeds this
value. Must be `0` or more. |
{.td-table--wide}
+> [!NOTE]
+> The `pd.balance-leader-*` settings are introduced in 1.8.0 (`master`
branch). With `pd.balance-leader-enable` turned on, PD periodically balances
partition leaders across Stores:
+> - A balance check is skipped if a shard group is unhealthy, lacks recent
heartbeats, is undergoing membership changes, or holds a shard on an inactive
Store.
+> - Planned transfers are only scheduled to Stores whose state is `Up` and
whose replication log is up to date (within 1000 applied entries of the leader).
Review Comment:
Thanks, fixed in cb41bf46b. The bullet now calls the 1000-entry limit an
estimate: PD works it out from Store heartbeats sent at different times, so a
burst of writes can hold back a caught-up target or let a lagging one through.
In the lagging case, writes to the partition can stall until the target catches
up or the transfer times out. The Chinese page says the same.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]