tsm server crash twice in the last 4 days

[rogwall]

Out TSM server crashed with a core file in /opt/tivoli/tsm/server/bin directory, the actlog just beore the crash were:



03/07/2006 11:20:12 ANR2017I Administrator admin issued command: select *

from client_Schedules where SCHEDULE_NAME='select

SCHEDULE_NAME from associations' (SESSION: 25800)

03/07/2006 11:20:12 ANR2034E SELECT: No match found using this criteria.

(SESSION: 25800)

03/07/2006 11:20:12 ANR2017I Administrator admin issued command: ROLLBACK

(SESSION: 25800)

03/07/2006 11:20:21 ANR2017I Administrator admin issued command: select *

from client_Schedules where SCHEDULE_NAME=(select

SCHEDULE_NAME from associations) (SESSION: 25800)

03/07/2006 11:20:21 ANR2954E An SQL subquery used in a scalar expression must

yield a table that contains at most one row. (SESSION:

25800)

03/07/2006 11:20:21 ANR2017I Administrator admin issued command: ROLLBACK

(SESSION: 25800)

03/07/2006 11:20:30 ANR2017I Administrator admin issued command: select *

from client_Schedules where SCHEDULE_NAME="(select

SCHEDULE_NAME from associations)" (SESSION: 25800)





I removed the core file and started the server and it came up fine. The DB available size is 167GB and it is 75.6% used.



It has happened second time in the last few days, on different commands. Please suggest. We never had this problem earlier, at least in the last 10 months since I am working on it.



Thanks,

:sad:

 

[chad_small]

Looks like there could be a problem with the DB. A bad SQL call caused it to crash? It could also be a bad record. Look for any errors in the DB that reference a bad DB record. If you don't see that look at the command and see if it needs editing. It could be a bug in the code when a bad SQL statement is passed. Any additional information would help. Is there an OS error for the crash? In AIX I usually look for any other problems (hardware usually) that could have caused it. Again to me it looks like a possible bad SQL call.

 

[stanleyb]

Are you running Server 5.2?



I think the last entry in your actlog gives it away and Chad is correct - it could be a code bug in the way tsm is handling an invalid sql query:



>>> select * from client_Schedules where SCHEDULE_NAME="(select SCHEDULE_NAME from associations)"



---i don't think those double quotes should be there---



I can always crash our 5.2 server by running a select statement inside quotes like you have there. If I run something as simple as "select * from nodes" (*INCLUDING* the quotes) I can crash the server every time. Someone else where I worked discovered this accidentally one day... so I tested it on a test server to make sure the quotes were the problem



I tried it recently on 5.3 and got a syntax error rather than a core dump.

 

[rogwall]

Yes we have server 5.2.2.0

Yes I ran exactly the same:

tsm> select * from client_Schedules where SCHEDULE_NAME="(select SCHEDULE_NAME from associations)"

No, there were no hardware/disk/etc errors on the box, we are using Solaris 8.



1. Just to eliminate Chad's second suspicion of possibility of error in DB, how can I check for bad records etc. ?

2. What will be correct syntax for the command I want to run (I want to list all the field of Client_Schedules that have some associations.



thank you very much.

 

[may]

just migrating the TSM server to 5.2.4 should resolve the Prob



have a look on the Aparts Fixed in this version

 

[may]

ho sorry I forget,



the correct SQL syntax:



select * from client_Schedules where SCHEDULE_NAME in (select SCHEDULE_NAME from associations)

 

[rogwall]

I want to use multiple where conditions, the second where condition is "where EXPIRATION is NOT NULL"...... I am reluctant to use the wrong quotes....