CORAL Server Meeting

Tuesday 26 Aug 2008, 17:00 → 18:00 Europe/Zurich

31/2-029 (CERN)

Design and development issues of the CORAL server project, and integration with the ATLAS dbproxy project. Phone conference details: Dial-in: +1-510-665-5437 Passcode Tuesday : 8083 Passcode Wednesday : 8082 Agenda: - Experiences with the latest CORAL server (CORALSERVER-0-1-3-preview) - Andy, Zsolt (10') - Development status: server, access plugin, proxy, tests - Andy, Alexander, Zsolt (15') - More results of the CoralServer profiling - Zsolt (15') - Test and monitoring tool progress - Alexander (5') - Plans for the next week - All (10')

Andrea's minutes from August 19 Tests with coral head - still broken (rw transaction starts later than ro transaction) - was probably tested only on server and not on cache (or is it that test changed?): check - make sure that we have tags also on tests Tests with SEALed version - not binary compatible after all - Andy tested 195, but expect that 197 should have the same problem - problem is that binary compatibility was designed in exceptions but not in the time stamps Performance tests (Reiner from discussion with Alex before he left) - both client and server are waiting - Zsolt: maybe this is because there is a 5s timeout - Reiner: puzzled that cache takes the same time, should be lower... => higher priority to understand the waiting rather than implementing caching (if caching was a problem, proxy should make a difference) Actions (on Zsolt) - fix sealed binary compat - fix rw/ro in head test - fix waiting times - try cache - document -------------------------------------------------------------------------------------------------- Andrea's minutes from August 26 Present: AK, ZM, AV, RB On the phone: AS Tests with SEALed version - latest code by Zsolt (1.3) solves binary compatibility in timestamp - but now Andy gets new different failures, the server crashes (this is a test with many clients simultaneous, did not try only one) (this is a test directly against the server, did not try proxy yet) - Andy will send the logs to Zsolt for analysis - Alex will implement a test to run several client threads Tests with coral head - new version of tests is being prepared (still need to figure out some technicalities about plugin loading) - not clear if transactions have been addressed (Alex did not change the relevant code) Performance tests - reminder, was 7 minutes cool test via server vs. 50s against db directly - Zsolt: instrumented code with monitoring info - currently analysing the results for integration tests - Zsolt: seems that schema operation takes the longest fraction (but then why does the proxy not improve the situation?) (but only executed integration tests, will try cool tests instead) - there does not seem to be any timeout involved - no timing difference if using same host or different hosts - Reiner: it must be waiting time (eg in select), must find out > Zsolt: there is no timeout in select > Reiner: strange, would guess there must be a timeout in this select - Zsolt will enhance the 'ping' test by sending dummy payload > 0 (configurable) and waiting some time on server (configurable) to emulate server side processing of relational data - unlikely that delay comes from logging either (tried /dev/null) Monitoring - Alex prepared a first prototype Priorities (according to Reiner) 1- SEALed crashes > Andy send the logs to Zsolt 2- deSEALed transactions (by the way, if now we run the same tests with and without SEAL, why do we not see a problem with transactions also in the SEALed tests?) > Andy should rerun the tests 3- performance > Zsolt and Alex should analyse