Hi Juergen,
your Indexer-0 log looks healthy. Is it possible that you ran the above commands multiple times to have it already be shutdown? If not, this may be another mystery to look into, but for the moment it seems to be inactive. You can verify this, as Chris said, via "bin/omero admin diagnostics".
helmerj wrote:Please find the Omero log file attached to this post
Looking into it. Thanks.
I have not been reusing sessions on my last import these log file have been generated with due to the fact that it was even slower than my initial approach starting each import session individually. It was not much slower but it definitely didn't speed things up. as I had to keep the session alive using omero session keepalive -f 500 it did complicate my system as that session would be alive indefinitely unless I killed it. So I abandoned that approach. I have it checked into a separate branch of my project so if you think you will be needing that feedback on this particular type of import, let me know and I can generate it for you.
Hmmm,...you shouldn't need to use keepalive manually, since the importer itself will call that. If there is nothing to import (not that that's a problem at the moment), then the session will time out and clean itself up. I was just trying to come up with a short shell snippet to do what you needed but ran into
ticket:4224. It may be necessary to provide you a short snippet of Python to do what you want. Roughly:
- Code: Select all
uuid = get_stored_uuid();
try:
check_uuid(uuid)
else:
uuid = get_new_uuid()
store_uuid(uuid)
return uuid
There's a short window in there, where you could be using two sessions, but that shouldn't be problematic.
Except for the fact that you couldn't use more than 4 sessions. Do you have any information about that?
I have created the postgres log using the auto_explain setting of my earlier post. Everything above 3ms is logged. Generated from a complete import of 1404 image files in 6 parallel import sessions.
I've attached some initial parsing of your logs. There doesn't seem to be too terribly much ome_nextval contention, so perhaps that's a red herring. The values seem evenly distributed across the rest of the application, i.e. inserts and selects, so perhaps we can try adding the attached indexes and compare the results.
As far as possible, we should try to isolate the tests:
* no OMERO indexer
* clean database
* run 1404 image in 6 parallel import sessions test
* save logs
* clean database
* add indexes
* run 1404 image in 6 parallel import sessions test
* save logs
Then we can compare the results. (Was the database clean before this run?)
One final question: do you have any logs from a single session'ed import? That might also be worthwhile to examine.
Thanks,
~J.