You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
cluster checks are now unconditional and have exponential backoff. previously they ran at a fixed interval only when there were no commands that could generate a MOVE/ASK error. That wasn't good enough for detecting newly added slaves
Copy file name to clipboardExpand all lines: README.md
+22-4Lines changed: 22 additions & 4 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -1358,11 +1358,29 @@ Additionally, `nchan_stub_status` data is also exposed as variables. These are a
1358
1358
legacy name: push_message_timeout
1359
1359
> Publisher configuration setting the length of time a message may be queued before it is considered expired. If you do not want messages to expire, set this to 0. Note that messages always expire from oldest to newest, so an older message may prevent a newer one with a shorter timeout from expiring. An Nginx variable can also be used to set the timeout dynamically.
1360
1360
1361
-
-**nchan_redis_cluster_check_interval**
1361
+
-**nchan_redis_cluster_check_interval_backoff**`<floating point> >= 0, ratio of current delay`
1362
1362
arguments: 1
1363
-
default: `5s`
1364
-
context: http, server, upstream, location
1365
-
> Send a CLUSTER INFO command to each connected Redis node to see if the cluster config epoch has changed. Sent only when in Cluster mode and if any other command that may result in a MOVE error has not been sent in the configured time.
1363
+
default: `2 (increase delay by 200% each try)`
1364
+
context: upstream
1365
+
> Add an exponentially increasing delay to the Redis cluster check interval. `Delay[n] = (Delay[n-1] + jitter) * (nchan_redis_cluster_check_interval_backoff + 1)`.
1366
+
1367
+
-**nchan_redis_cluster_check_interval_jitter**`<floating point> >= 0, (0 to disable)`
1368
+
arguments: 1
1369
+
default: `0.2 (20% of inverval value)`
1370
+
context: upstream
1371
+
> Introduce random jitter to Redis cluster chck interval, where the range is `±(cluster_check_interval * nchan_redis_cluster_check_interval_jitter) / 2`.
1372
+
1373
+
-**nchan_redis_cluster_check_interval_max**`<time> (0 to disable)`
1374
+
arguments: 1
1375
+
default: `30s`
1376
+
context: upstream
1377
+
> Maximum Redis cluster check interval after backoff and jitter.
info: "Send a CLUSTER INFO command to each connected Redis node to see if the cluster config epoch has changed. Sent only when in Cluster mode and if any other command that may result in a MOVE error has not been sent in the configured time."
info: "Introduce random jitter to Redis cluster chck interval, where the range is `±(cluster_check_interval * nchan_redis_cluster_check_interval_jitter) / 2`."
0 commit comments