CORAL Server Meeting

Tuesday 11 Nov 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: - CORAL server protocol review (25') - Experiences with the latest CORAL server (15') - Development status: server, access plugin, proxy (10') - Plans for the next week - All (5') - AOB - All (5')

Paper CoralServerProtocol.pdf

Next week plans: - Modify the protocol description: - In case of segmentation error, no error recovery on client side, the server discards the whole message, the error message sent back is unchanged - Only the client closes the session (clarify the text) - CORAL server 0.1.11 plans: - synch with the change above - Test around the AttributeList marshaling, to figure out what happened in the failing HLT test case - Isolate the ORA error reported earlier (the CLOB reading one) - Test bed: - Finalize the first version of the cycler - Set up a separate database server for the stress tests (not to interfere with the nightlies) - CORAL - Support for ATLAS integrators on the CORAL documentation side - Write the NSS article - Continue with the CORAL server design document -------------------------------------------------------------------------------------------------- ALexander's transcript: Present: AK, ZM Over the Phone: AS, MW (Martin Wache) 1.Topic, Comments to the protocol: AS: checked segmentation and don't understand segmentation procedures and error handling AS: if the packet is out of order, what does the client send back? ZM: segmentation error will be sent only from the server, if an error occours discusion went into details on segmentation AS: what happens if one package doesn't reach the server ZM: server always opens a buffer for each package to store the segments in it ZM: the propper packets are in the buffer, faulty ones are dropped ZM: up to the client how to handle segmentation errors of packets AS: use case : one package is missing in the middle ZM: the segmentation number of the last correct package will be sent back to the client AK: segments are not in correct order ZM: client will send the whole massage again AK: how is the situation in the moment ZM: at the moment there are no segmentated messages AS: when client side is out of order solution: AS: server should expect the complete package, that is the simplest way at the moment AS: if an error happens, the client will throw an exepction security of the password and username: AS: password and username in clear-text is an security issue ZM: we can use a secure channel, but in general secure discusion later after functionality AK: we can add a simple function to encrypt password and username with a known key for both sides 2. Topic, Development status: AS: main interest is the schema issue, reported in the mail before answer of the questions in the mail: AS: the used schema is a regular one, no special atlas database possible reasons: ZM: problem in the coral messaging, converting CLOB AK: maybe a wrong order in the Attribute list focus changed to overall feeling on the current server version: AS: everything compiles fine, no failures to report AK: we need general tests for type checking to prevent issues as above focus now on new implementations/developments ZM: nothing happens in this week on the server side AK: some improvements on the monitoring tool (backtrace of seg faults, colored display) AK: simple cycler for night long testing the server continued with proposals: ZM: we need separate schemas for those tests, can stack the db and other users AK: log parsing in version 1_9_5 ZM: better to have two versions of filtering the log (threadid, taskid) ZM: providing a threadid in version 1_9_5 is not possible 3. Topic, plans for next week: ZM: no time for coding, need to write documentation ZM: try to solve the errors on schema ZM: new boost version (1.3.6) available, but no features used AK: implementing the complete cycler MW: first, can read the documentation to understand and confirm it as not involved person