Requeue Failed cachestore write operation.

Question

I am trying to requeue failed cachestore write operation , for this I add <write-requeue-threshold> property in coherence-config file,but I am not sure about its behaviour
1) what will happen if number of cachestore failed operation are greater than threshold specified, means will the extra objects also be requeued of simply dropped off?
2) is the sequence of write operation maintained, let say
i) DB is down at time T1
ii)Object A is requeued at time T2
iii)DB is up again at time T3 and at the same time another object B is also placed in queue
then will object A be written first or Object B.
Gaurav

Answer 1

user8157603 wrote:
I am trying to requeue failed cachestore write operation , for this I add <write-requeue-threshold> property in coherence-config file,but I am not sure about its behaviour
1) what will happen if number of cachestore failed operation are greater than threshold specified, means will the extra objects also be requeued of simply dropped off?
AFAIK, if the number of entries which failed to be written grows above the write-requeue-threshold, then the surplus entries will just not be queued for retries. Coherence would just not have the information that it should write it out that entry because it is still dirty.
2) is the sequence of write operation maintained, let say
i) DB is down at time T1
ii)Object A is requeued at time T2
iii)DB is up again at time T3 and at the same time another object B is also placed in queue
then will object A be written first or Object B.
If you mean object A and B are put in the cache for the same key, then, one of the following two scenarios can happen:
As there is no such thing as two events happening at the same time when race conditions are guarded against, let's say that the cache processes the placement of Object B into the cache at time T4 which is very close to T3. The question is how does T4 compare to the time when the write-behind thread starts collecting entries to write and pass it to the cache-store (T5) .
If T5 < T3 and T5 < T4 then the database is still down and the write does not succeed, and afterwards Object B will be enqueued for its first write, so the entry is not retried yet. Object A will not be written to the DB at all.
If T3 < T5 < T4 then Object A will be written and Object B will be enqueued for writing in the next batch write.
If T5 > T4 then Object A will not be found by the write-behind thread at all, and Object B will be written out.
Best regards,
Robert