03-20-08
--------

Summary
- Bug fixes 
- Preparation for export
- Export SU
- pgsql function drms_series() acts up on DB jsoc_test
- Thoughts on why slony log shipping can run alone

Bug fixes
---------

1. create_series hangs upon permission problem.

Bug due to missing a db_unlock() in
db_postgresql.c. drms_server_abort() needs a db lock to rollback. When
it can't get it, it waits, hence the hanging problem. I generated this
bug from last version.

2. fixed bug in drms_env.c drms_free_env() to handle early bailout
when there is no namespace. sum_inbox and sum_outbox may not have been
initialized.

preparation for export
----------------------

a. Link /SUM0-8 on jsoc:/web/jsoc/htdocs
   cd /web/jsoc/htdocs; ln -s /SUM? .
   - Prevent directory listing in SUMS directory
     cat /dev/null > /SUM?/noindex.html
     added noindex in httpd.conf
     DirectoryIndex index.html index.html.var index.php noindex.html
     (checked with Brian, there might be better solutions)
   The level below /SUM?/ can still list file names

b. correct permssion for jsoc.export
revoke select on exports from public;
revoke select on exports_seq from public;
grant select, insert, update, delete on exports to sdodata;
grant select, update on exports_seq to sdsdata;

group sdsdata has full access to jsoc.exports

Export SU
---------

Working on a perl script to export a SU. This will serve both remote
SUMS and IDL clients. The latter may get the fits header of a record
from DRMS database in a JSON format along with a SU. This approach
eliminates the assembly step on our end which is well suited for IDL
usage. 

pgsql function drms_series() acts up on DB jsoc_test
----------------------------------------------------

drms_series() function failed because schema change in drms_series
template, jsoc_test uses su_karen.drms_series, jsoc uses
hmi_ground.drms_series. 
schema change: added dbidx 
- manually using alter table
- masterlists.c added a line in create table. 
the order of the columns looks different from \d

I can't repeat the problem on jsoc. I create a new namespace
karen_test. no problem there. I will keep an eye on it.

Thoughts on why slony log shipping can't run alone
-------------------------------------------------

The replication master and slave run in lock step, i.e., until the
slave reports that it has already processed certain replication
events, the master can't remove such event from its internal log. In
log shipping mode, there is no feedback from the slave to the
master. The master has no hope of reducing its event log.