Please sign in to comment.
HPCC-8189 JobQueue updating dali too frequently
Both eclccserver and roxie were setting short timeouts on the call to jobq.dequeue, in order to ensure they could close down cleanly and quickly. However, this resulted in frequent writes to dali to update the queue status (several per second), meaning dali would never be idle and the inc file would grow unnecessarily large. Use the async cancelAcceptConversation method means we can close down even faster, and avoids the dali noise. There is still a timeout every 900s (in a defualt configuration) on the dfuserver monitor queue. That cannot be eliminated without more significant refactoring. Signed-off-by: Richard Chapman <firstname.lastname@example.org>
- Loading branch information...
Showing with 27 additions and 7 deletions.