On Thu, Apr 25, 2013 at 11:28:41AM +0200, Pavel Březina wrote:
Here is design document for #1713.
https://fedorahosted.org/sssd/wiki/DesignDocs/PeriodicalRefreshOfExpiredE...
It is not yet fully complete because I believe is miss one thing.
The whole point of this ticket to to reduce latency when refreshing
a large amount of entries. So we create a periodic task that will
refresh the entries out of bound instead of doing it during login
time.
But I'm not sure if this is enough, because the amount of entries
may be to large to refresh all of them in reasonable time. It may
also block the back end for a while. I think we should do this in
some sort o batch.
Jakub suggested that we will adapt to the current network status.
For example:
1. try to update 200 entries
2. if the update finishes in less than 5 seconds, increase the batch
3. if it finishes in more than 5 seconds, decrease the batch
4. next batch will be filed in the next period
I think this is what we discussed on the phone some time ago. The
numbers were just pulled out of thin air.
Another approach may be to make the batch fixed size configurable in
sssd.conf.
Suggestions are welcome.
_______________________________________________
sssd-devel mailing list
sssd-devel(a)lists.fedorahosted.org
https://lists.fedorahosted.org/mailman/listinfo/sssd-devel