Description of problem: just transferring a few bytes of translations takes 10~20 seconds. it's too slow and takes the time too long when pushing multiple documents.
Mind telling us what projects you were pushing and what command did you use for pushing? It is possible you were experiencing Bug 1102964, which is fixed in 3.4.2. I hereby close this bug as duplicate of Bug 1102964. Please reopen this bug if same thing happened again for Zanata server version after 3.4.2. *** This bug has been marked as a duplicate of bug 1102964 ***
(In reply to Ding-Yi Chen from comment #1) > Mind telling us what projects you were pushing and what command did you use > for pushing? man-pages-l10n > It is possible you were experiencing Bug 1102964, which is fixed in 3.4.2. There are no way to try 3.4.2 at this moment since the instance on zanata.org is still old one apparently. though trying with --copy-trans false is even slow. Aside from that pulling in translations is slow too.
(In reply to Akira TAGOH from comment #2) > (In reply to Ding-Yi Chen from comment #1) > > Mind telling us what projects you were pushing and what command did you use > > for pushing? > > man-pages-l10n > > > It is possible you were experiencing Bug 1102964, which is fixed in 3.4.2. > > There are no way to try 3.4.2 at this moment since the instance on > zanata.org is still old one apparently. though trying with --copy-trans > false is even slow. > > Aside from that pulling in translations is slow too. This is what I see in https://translate.zanata.org/zanata/ Zanata 3.4.2 (git-server-3.4.2-dirty) What version did you see? I will investigate the pull part.
Ah, I thought that meant the API version. [INFO] client API version: 3.3.1, server API version: 3.3.1
and tried the above "--copy-trans false" thing on that version.
Guess it is different problem then, reopen.
Some notes from our estimation discussions: Processing on the server is the main time taken. Network transfer is not the bottleneck. With import = merge, we have to check history. Might be able to improve this query (i.e. load history in memory rather than query database each time). Issues started after we moved to OpenShift - could be related to async and threading performance on OpenShift. We'll take a time-boxed approach, investigate and fix what we can in the allocated time.
Topics of interest: - Pushing of small files - merge=import history search - Web sockets for eap 6.3 - HornetQ long polling
https://github.com/zanata/zanata-server/pull/571
See: https://github.com/zanata/zanata-server/pull/571#issuecomment-55547577
On my machine: man-pages-l10n tooks 86.11s user 8.43s system 3% cpu 49:45.66 total with command: time mvn zanata:push -Dzanata.pushType=both -Dzanata.srcDir=pot -Dzanata.transDir=po man-pages-l10n contains 69849 text flows (entry), 0.04 was translated, so it is about 72642.96 strings to be processed. So the speed is 24.33 strings/sec. Much better then previous 3 strings/sec. Zanata 3.5.0-SNAPSHOT (git-jenkins-zanata-server-github-pull-requests-2090)