keith-turner commented on issue #3666:
URL: https://github.com/apache/accumulo/issues/3666#issuecomment-1659213338
The approach I am working on at the moment is making the event coordinator
accept different types of events so that the TGW can respond accordingly.
Currently have the following events types.
* ALL : an event that impacts all datalevels off accumulo. Like a tserver
dying. This kind of event will make all the TGWs do a full scan.
* DATALEVEL : An event that impacts an entire datalevel. This happens when
the TGW has made changes and wants to immediately run again. This type of
event will only make one TGW do a full scan.
* TABLE: An even that impacts an entire table. This happens after
something like table creation. Only one TGW will respond to this event and it
may only scan the data for that table.
* TABLE_RANGE : An event that impacts a range within a table. This happens
with things like compaction, split, and on demand hosting request. Only one TGW
will respond to this event and it may only scan the data in a that range of the
table.
I am working to make the TGW have different behavior depending on the type
events that have happened and the time since it last full scan. So if a
TABLE_RANGE event comes in and its only been a few seconds since the last full
scan, then the TGW will just handle the tablets in that range and not do
another full scan.
When the TGW does do a full scan, still want the behavior introduced in
#3447 where the more targeted events above are injected for immediate
processing.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
