Affects Version/s: 10.10, 11.1-SNAPSHOT
NXP-25312 (Nuxeo 10.3) computations have a retry policy.
The policy for the audit log writer computation is:
Which means 3 retries with 1s exponential backoff delay up to 10s delay, so delays are 1, 2 and 4 seconds, or:
- t: failure
- t+1s: retry 1
- t+3s: retry 2
- t+7s: retry 3
With this configuration the tolerance is 7 seconds shortage, after this, the processor stop and a manual restart is required to resume activity.
There is no good reason to not tolerate a 15min shortage by default.
This could be done like this:
time between retries for the 10 first retries
1, 2, 4, 8, 16, 32, 60, 60, 60, 60 -> 5min05
then 10*60 -> 10min
The Elasticsearch re-index bulk action can also benefit to use this retry policy to be able to support Elasticsearch failure.
Note that for Nuxeo 9.10 there is no retry policy mechanism, a computation in failure stops the processing and requires a manual restart.