...
Tip | ||
---|---|---|
| ||
The Operation Timeout, Retry Interval Time, and Max Number Of Retries settings together with the Advanced tab setting mz.cb.lock.timeout.in.secs, work jointly. To understand how, see the following explanation. If you for example use the lookup function with transaction, a LOCK ERROR means that the lookup failed because there was already a transaction lock held on the object that the lookup was trying to get from Couchbase. That is, the lookup was not able to get a lock on the object. What is most important here is how the communication between and and Couchbase works. In case you use the lookup function it works like this:
This is why the Operation Timeout parameter does not play a crucial role here. The Operation Timeout parameter is only used when Couchbase is stalling and does not send an answer for a long time. Assuming that the object is locked and the system received answer with LOCK_ERROR, it will wait for the amont of time specified in Retry Interval Time and then retry the request towards Couchbase. This process is repeated until Max Number Of Retries is exhausted or the answer contains a successful result code and the object is retrieved from Couchbase. So, if you for example set Retry Interval Time to 100 ms and Max Number of Retries to 10, you get 100 ms * 10 retries = 1000 ms. However, from a MZ perspective, the timeout of 1000 ms is just for a single lookup attempt and not for the whole lookup operation. To clarify, if you get a failed lookupMany with LOCK_ERROR in the system log, it means that the lookup failed because it tried to get an object with a lock but failed after Max Number Of Retries. That is, some other process (or thread) was holding a lock on that particular object for longer than 1000 ms. This is possible if the Advanced tab setting mz.cb.lock.timeout.in.secs is larger than 1000 ms. In other words, you need to decide what is more important - that the operation succeeds or that the operation is fast. |
...