JRA1-UK

Weekly Meeting

by Steve Fisher

→ Europe/Zurich
CR04

CR04

Description
<html> <head> <meta name="generator" content= "HTML Tidy for Linux/x86 (vers 1st March 2002), see www.w3.org"> <title></title> </head> <body>
  • Reports on Meetings
    • Design meeting in ANL - Steve F
    • INFOD - Abdeslem/Steve F
  • Actions
    • Organise review of spec - Steve F - 11/7/2005
      1. Views: are they going to be read-only? If so, I'll make sure it's clear in the spec where view names are allowed in place of table names, and where they're not.
      2. Version numbering: no-one else has commented yet on Martin's discussion of this problem.
      3. Error numbers: down to John, but it will take a lot of work (maybe too much for gLite 1.5).
      4. List of properties for set/getProperty: I could make a guess...
      5. Complex Query definition: sooner or later our parser is going to have to validate SQL SELECTs properly, and we're going to have to test it properly. For now, it might just have to stay as a fixme. Simple queries are now fully defined.
    • Review Registry Replication and how to introduce it - Steve F, Rob and John - 11/7/2005
    • Deploy new GIN on Prototype - one more CE to be done - Alastair - 25/7/2005
    • Run the lxgate24 checking script - we may need to remove this action - Antony - 25/7/2005
    • Check that list of features for immediate deprecation appears in gLite 1.3 release notes - Alastair - 1/8/2005
    • Write test specification. - John - 1/8/2005
    • Finalise "Sensible use of R-GMA" note - Steve F - 1/8/2005
    • Investigate SFT2 failures (timeout) - Antony/Alastair - 8/8/2005.
    • Ask Piotr to change SFT2 to display LCG version - Antony - 15/8/2005
    • Set up RGMA19 with LCG2_6_0 YAIM installed code and the SA1 job monitoring archiver. - Alastair - 15/8/2005.
    • C API test by end of 15/8/2005. Note that what is not done today will be deferred to allow work to continue on the spec review. - John - 15/8/2005
    • C++ API test be end of 19/8/2005 - Martin - 15/8/2005
    • Python API test by end of 19/8/2005 - Rob - 15/8/2005
    • Spec issue list by 17/8/2005 - John - 15/8/2005
  • LCG2_4_0/2_6_0 - Antony
  • Status of patches for R-GMA and APEL - Alastair
  • R1.3 and prototype - Alastair
  • Bugs
    • 4438 - R-GMA doesn't check types - Steve H
    • 6512 - More parsing errors - Steve H
    • 7283 - Inserting ''s - Steve H
    • 7332 - Services such as rgma-servicetool are not services. - Martin
    • 7442 - Remove redundant tables - Antony
    • 7452 - Inconsistent use of stdout and stderr in java API - Steve H
    • 7676 - Need to find out server limits on parameters - John
    • 7704 - Publication rate and LRP for site and service publishers should be configurable - Martin
    • 8099 - Archivers are inconsistent - Martin
    • 9109 - API is missing a "show tables" equivalent - John
    • 9272 - tomcat log4j warnings - Martin
    • 9523 - StreamProducer doesn't crop VARCHARs to size defined in Schema - Steve H
    • 9665 - Cpp API makes a new connection for each server call resulting in slow performance - Alastair
    • 9674 - org.glite.service-discovery-rgma.java has description not set - Steve H
    • 9699 - ArchiverThread doesn't check column types when inserting. - Steve H
    • 10066 - Reassigned item: Site Archiver - ???
  • Remind Bugs
    • 5703 - case sensitive varchar's when creating an archiver
    • 6288 - HTTP proxy support is inconsistent
    • 6475 - Case sensitivity for table names is problematic
    • 6493 - Bad case sensitivity for insert
    • 6496 - Bad/missing error messages with select
    • 6510 - Case sensitivity depends on query type
    • 6514 - Case sensitivity error with a secondary producer
    • 7149 - Formatting strings for output
    • 7188 - Bad messages if a port isn't open
    • 7189 - How do you know the primary key?
    • 7220 - Bad error message with ssl error
    • 7225 - All client interactions should go via the local servlet box
    • 7226 - rgma client configuration succeds, but a client test fails
    • 7276 - Bad error message for expired proxy
    • 7384 - Proxy should default to the standard location
    • 7399 - API errors with expired proxy
    • 7542 - Overriding the rgma.conf
    • 7635 - Problems with registry configuration
    • 7654 - Detection of an error with TRUSTFILE
    • 7726 - Bad error message in web browser
    • 8992 - browser failed to update column definitions
    • 9395 - Registry API causes deadlock
  • gLite 1.5
    • Definining the contents of 1.5
      1. Are time intervals on continuous, latest and history queries supported?
      2. Are we still having to support non-chunking producers?
      3. Are we adding the new metadata columns in CREATE TABLE?
      4. Are metadata columns being handled properly by all three types of producer?
      5. Are the old metadata columns (MeasurementDate, MeasurementTime) withdrawn?
      6. Do we have separate latest and history tuple stores?
      7. Are the LRP and HRP implemented as specified (i.e. are tuples retained until they expire and removed when they expire; are latest queries limited to un-expired tuples only)?
      8. Is the close() logic of Primary Producers (wait for history tuples to expire and keep registered until they do) implemented? What about destroy()?
      9. Is the close() logic of Secondary Producers, On-demand Producers, Consumers (destroy immediately) implemented? What about destroy()?
      10. Is the mediator service implemented as a service in its own right? Answer: no, it's currently internal to the Consumer Service.
      11. Are we using SOAP for the interfaces yet? Answer: no, we have an equivalent servlets interface defined by Steve's doc.
      12. Do they services send showResourceSignOfLife() messages to the registry on behalf of producers and consumers, and is this now its own thread?
      13. Has the Perl API been withdraw?
      14. Does getVersion() return anything yet?
      15. Have we implemented authorization, or do we still just have authentication and host filtering (access control list)?
      16. Is Registry replication implemented?
      17. Is Schema replication implemented? Answer: no.
      18. Are Registry and Schema services able to act as proxies? Do they now handle the failover from one replica to another or is it still in the API?
      19. Are createRegistry/Schema and destroyRegistry/Schema implemented?
      20. Are (logically named) permanent tuple stores implemented? Are old-style user-named tuple stores still supported?
      21. Do we still have more work to do on queues and timeouts for remote calls?
      22. Is RGMASecurityException implemented yet?
      23. Are error numbers implemented yet? Answer: no.
      24. Is the lookup file mapping registry/schema+vdb to URL implemented yet and is their a mechanism for updating dynamically, or is it static for now?
      25. Are all valid combinations of query type and storage now supported (e.g. latest/memory and continuous/database)?
      26. Are VDB prefixes on table names supported everywhere? Are they tolerated? Are multi-VDB queries supported by the mediator? Are VDB-union queries supported?
      27. Is all user-defined SQL now validated?
      28. Is user data checked against the table definition? Do we still quietly convert/truncate data? Are all the new types (and just them) supported?
      29. Are getProperty/setProperty implemented? Have they replace *get*Status? What are the implications for scripts that used to call these directly? Partial answer: we need to change the scripts...
      30. Are the new system interfaces fully supported? Are we deprecating all of the old ones in 1.5, or will some things have to wait longer?
      31. Is ping implemented yet?
      32. Do we support table/column name mapping in producers? Do we need to or does MySQL support everything we'd now allow anyway? Case sensitivity?
      33. Are SQL keywords treated as case insensitive everywhere in R-GMA?
      34. Do Primary Producers successfully block inserts when their tuple stores fill up, or do they just fall over?
      35. Do Primary and Secondary producers send HRPs to the Registry and if so, do they update them in showResourceSignOfLife to match the data they actually have available? Does the mediator use this information yet?
      36. Does insertList now try to insert ALL tuples before failing if one fails?
      37. Can Secondary Producers answer continuous queries?
      38. Have we removed server-side limits on LRP, server-side minimum on HRP and added server-side maximum and minimum (which may be identical) for TI?
      39. Is the QoS attribute implemented (albeit unused)?
      40. Do SP's now attempt to recreate any of their consumers that fail?
      41. Has HTTP support been withdrawn from the ODP? Does it still return result sets unchecked from query handler to consumer service? Does it support chunking?
      42. Does SQL SELECT support for simple queries match the spec?
      43. Is startDirected implemented yet?
      44. Is the endOfResultsFlag on a result set implemented yet? For all types of query?
      45. Is createSecondaryConsumer implemented yet?
      46. Are addProducer and removeProducer now treated as notifications?
      47. Is encrypted streaming implemented?
      48. Are getAllProducersForTable() and getMatchingProducersForTables() implemented as described in the spec?
      49. Does registerContinuousConsumer() still do mediation and return plans, or is it now in line with the specification?
      50. Does an entry timing out of the registry still generate a removeProducer?
      51. Which of createTable/dropTable, createView/dropView, createIndex/dropIndex actually do anything? Is tableAuthz?
      52. Do Primary Producers implemented indexes yet? Including RgmaTimestamp?
      53. Which of getAllTables, getTableDefinition, getTableIndexes, setAuthorizationRule, getAuthorizationRule, getFullTableDetails actually do anything? Do they match the spec?
      54. Does the mediator return all available (ODP) producers for a static query? Does the consumer correctly contact all of them?
      55. Are we imposing the new table name rules? Do we allow all new data types in CREATE TABLE (i.e. adding DOUBLE PRECISION, TIMESTAMP(n) and CHAR(n)) and in a result set? Do we require "DATE/TIME/TIMESTAMP" prefixes on literals of those types? Do we tolerate it?
    • Writing it
    • Testing it - where do we put most effort - presumably the new stuff
      • API/Java/ServiceCalls
      • API/Java/ServiceExceptions
      • API/Java/ParameterChecks
      • API/C/ServiceCalls - John
      • API/C/ServiceExceptions - John
      • API/C/ParameterChecks - John
      • API/Cpp/ServiceCalls - Martin
      • API/Cpp/ServiceExceptions - Martin
      • API/Cpp/ParameterChecks - Martin
      • API/Python/ServiceCalls - Rob
      • API/Python/ServiceExceptions - Rob
      • API/Python/ParameterChecks - Rob
      • Service Test Group
      • Function Test Group
      • Simple Full System Test
      • Distributed Full System Test
      • Resilience Test
  • AOB
</body> </html>
actionlist
document
minutes