JRA1-UK Action List
Low Priority
- Design and code mediator - ??? - 27/4/2006
There is a problem with views. Need to get back the underlying table as
well as its description when querying something which turns out to be a
view rather than a table. Needs checking by someone.
- Code new streaming - ??? - 6/6/2006
Mostly done on consumer side. Needs checking by someone.
- Tuple stores - ??? - 12/6/2006
Coming along nicely
- Complete PP design and code it
- OGSA Naming into GLUE 2 - Antony - 2/10/2006
Wont be done till OGF.
- Follow up on gEclipse - Steve - 2/10/2006
- Having confirmed that it does not introduce compatibility
problems, make the exceptions explicit in each language where it can be
done without negative impact. This will only be done on the new code -
??? - 5/1/2007
- Ensure that C and C++ APIs have logging code IFDEFed so
that a developer can turn on logging for himself but deployed code has
no dependencies - ??? - 1/1/2007
- Consider changing the consumer to make use of "Tuple
Stores" It would reduce the volume of code though probably at the
expense of some performance loss. The "Tuple Store" is able to deal
with both memory and disk based stores and may make the SP trivial. -
??? - 15/2/2007
- Devise a way to conceal the DB user name and password -
??? - 21/2/2007
Medium Priority
- Follow-up on MonAMI - Antony - 5/7/2006
Only runs on insecure port. Discussion with Paul underway.
- Should SAM publish to GOC Node? Maybe Dave K could
publish.
- It is possible to access our services on port 8080 if the
server.xml has connector. We should be able to secure the service
however the connector is setup (as others may be using tomcat).
- New "simpler" init.d scripts for the flexi
- Try to break JDK socket library
Various ways to get socket into closed wait state. 1.4 and 5 look
similar. It would appear that JVM being out of memory can provoke this.
- Investigate protection against running out of
memory
He thinks that we should detect and restart the service - not clear how
to do it. Lots of discussion... Maybe we should make some measurements.
Alastair has done some work (it stops the service, trys to start the GC
and restarts the service). Tested with big linked list on 1.5 - it did
not recover well. Will add an RGMABusy exception to allow service to
say it is too busy. Maybe do a better job of estimating total memory
used.
- Service Proxy Module - add to GIN etc
Antony has done this for the job status raw archiver - can it be done
in a more generic way? GIN has some bug fixes to make as well. Need to
assign bug to Alastair - Antony will check that it is in Savannah.
- Consider load on registry in new design - consider stored
procedures in RDBMS
- Investigate Tomcat long running test (memory leak) and use
of log4j
Made contact with Torsten.
- Need to generate RGMABusy exceptions at the right time on
the HEAD server code
First need to define waht it is to "be busy".
- Eliminate the spec fixmes - Steve - 9/1/2007
- Dependency diagram in design needs updating - Steve -
15/1/2007
- Plan migration from physical DB names to logical - ??? -
21/2/2007
- Talk to Philipa next week about maintenance of registry
XML - Antony and Alastair - 1/2/2007
Ask again - but not before end of March
- Eliminate log4j
- Eliminate log4cxx/pp
- Use jaxp
- Move op authz to a filter on the
server - rather than
the service
High Priority
Alastair
- Client patch
patch
#1086 is
now with John Walsh. The test tar file and instructions have
been sent to John. It seems the RPMs have not been built yet
- Server patch
Tweaked test a bit to make it easy for John Walsh - should be submitted
today
- GIN patch
Tagged code - Antony will provide info on how to test then the patch
can be created
- Evaluate effort to do the local side of the registry.
- Upgrade machines
Memory for Parminder
- Define configuration for mon box and get it going
Work has started
- Cleanup ETICS stuff
- Tag monitoring tools
Created configuration on HEAD but needs to be branched
- Prepare configuration and building of new code before the
code is finalised
Will probably use 2 war files.
- Find why Alessandra's mon box is unhappy -
15/1/2007
CERN monb001 is also not happy. Threads and memory etc run
out. Runs out of memory in about 90 minutes.
- Code Registry - 27/4/2006
Registry is just about there - but without replication of course. Uses
a dummy schema. Using Martin's remote registry service which needed
some changes. Not yet described in design. It must also be protected
from bad ACL files. Will remove context in Registry and use
ThreadLocal.
Antony
- Get UI info so that we can get back in the RTM
Now waiting for wms #24148 - A new tag of wmproxy is going to
be released in the next days. Meanwhile Antony is
looking at the log file parser. New parser written. An upgrade
plan will be circulated and the new parser sent to Laurence.
- Watch bug on job publishing tables with VO and
predicate
22869 - Antony will agitate. Another mail sent but no response.
- #24643 Tomcat
can fail in various ways detectable by parsing its log file
- Find a way to look for GIN/ServiceStatus problems in R-GMA
so that we are useful to the infor supermarket.
- #22205
Upgrade
to Glue schema 1.3
- Write an R-GMA-probe or PrimaryProducer-probe https://twiki.cern.ch/twiki/bin/view/LCG/GridMonitoringProbeStandard
- #24641
LD_ASSUME_KERNEL no longer needed
- #24007 org.glite.rgma.flexible-archiver
contains deprecated calls
- #24029
org.glite.rgma.servicetool
contains deprecated calls
- #22167
GLUE
archiver has problems with stale lock files
Kush
- Fix bug in HEAD in ServletResponseWriter
Source should include stack trace, truncated if necessary. Now decided
that should not send the stack trace but just a simple message.
- Code consumer and task q monitoring service details
following recent e-mails and in line with WLCG monitoring requirements
Need to consider whether to use key value pairs in the response
or XML and look at the WLCG document. Need to also
update the inspector to match. In progress.
- Add logging to consumer in "approved" style (still log4j)
Ming
- Code Schema - 8/1/2007
Now working to use the TaskQ
- Add logging to schema in "approved" style (still log4j) -
13/2/2007
- #19645 The
Sub exception org.edg.info.ReadTimedOutException is an inappropriate
type
Parminder
- Convert tests to python, introduce more
automation and bring the document and the code in line with each other.
This will all be done initially on the 1.5 branch. One subsystem to be
completed first.
- Rule of thumb for how many mon boxes a site should run
- Ensure that the start call to Producer being added by
Alastair is tested and the 2 new exceptions added by Ming are tested.
Will be DONE once tested on HEAD.
Paventhan
- BDII with C++ with SAGA query style
Written code to generate the AST. Now working on the BDII
query string. Can generate filter string for BDII
queries. Getter methods added. The ListServices function needs
completing.
Steve
- Assign people to address migration from 1.5 to HEAD of
issues collected by Parminder
- Review chapter 2 of INFOD spec
- Write SAGA SD spec
- Improve release plan
- Add new new calls to list available VDBs to spec and
design
- Add to the spec that both client and mon box names
should be a stringified numeric IP addresses
- Add to the spec that both client and mon box names
should be a stringified numeric IP addresses
- Update design for schema following e-mail
May need to think again about the IndexToColumn table
- Add consumer and task q monitoring service details to
the design doc