Articles in this section

Why do I see batch timeouts under XDR peak load?

Detail


The RW internal Transaction queue is used by many internal transactions. (Batch-index, XDR, Re-Replication, background ops...) 

 

Answer

These transactions (XDR, Batch-index, Background OPS) share the same internal transaction queues (one per service thread).

In looking at the impact of XDR on batch-index, If XDR filled too many requests in the queue, batch requests will be interleaved with the xdr requests and can get delayed and sometimes hitting a timeout.

Records which timed-out (early timeout) in the RW Transaction Queue will NOT be re-queued for re-replication and can be the cause for missing XDR shipping of a few records.

The tuning of XDR
max-used-service-threads is also important. Having less max-used-service-threads would force all XDR reads to go through fewer internal transaction queues and get overloaded which would  impact other transactions using those same service-threads  (This could include the batch transactions).

Applies To Earliest Version

5.0

Applies To Latest Version

Current Version
Was this article helpful?
0 out of 0 found this helpful