on 06-22-2015 11:53 AM
Hello experts,
today I've experienced an error with indexserver not starting due to a large number of transactions in the sys.m_dev_rollback_transaction_table
This is described in SAP Note 2099859 and also in the following post:
I have performed steps to increase the stack size which at least allows me to start indexserver again. However the sys.m_dev_rollback_transaction_table contains a large number of values (129.087) and stays this way. The note suggests that after this value has decreased to zero I should reset the stack size parameters, but the value does not decrease by itself.
I am not sure I should leave the system in this state and I am looking for a way to get this value down. Are there any steps I could take to do this?
Regards, Dirk
If the entries remain, something must still be wrong (e.g. an offline service that is not removed from topology). Can you execute the following command and provide me with the results?
SELECT
VOLUME_ID,
HOST,
PORT,
COUNT(*) NUM_ENTRIES
FROM
SYS.M_DEV_ROLLBACK_TRANSACTION_TABLE
GROUP BY
VOLUME_ID,
HOST,
PORT
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
I assume that host and port are the expected values for the used host and the indexserver? In this case I would recommend you to check the indexserver trace file for errors that could explain the remaining entries. Can you furthermore execute the following command, so that we can get a better feeling for the entries?
SELECT TOP 10 * FROM SYS.M_DEV_ROLLBACK_TRANSACTION_TABLE
Host and port are expected. I will check the indexserver traces for any errors.
Query results are the following:
VOLUME_ID;TID;TYPE;PARTICIPANTS;HOST;PORT
3;80.673;PRERESTART;;vhcalhdbdb;30.003
3;80.674;PRERESTART;;vhcalhdbdb;30.003
3;80.675;PRERESTART;;vhcalhdbdb;30.003
3;80.676;PRERESTART;;vhcalhdbdb;30.003
3;80.677;PRERESTART;;vhcalhdbdb;30.003
3;80.678;PRERESTART;;vhcalhdbdb;30.003
3;80.679;PRERESTART;;vhcalhdbdb;30.003
3;80.680;PRERESTART;;vhcalhdbdb;30.003
3;80.681;PRERESTART;;vhcalhdbdb;30.003
3;80.682;PRERESTART;;vhcalhdbdb;30.003
I would be happy to look into it in more detail. If you open a SAP incident (including SAPGUI connection and logon data) and provide me with the incident number, I can take over and have a look at the problem myself. Optimally the SAP HANA user configured for transaction DBACOCKPIT has the SAP_INTERNAL_HANA_SUPPORT role assigned, so that it can access the internal M_DEV_ROLLBACK_TRANSACTION_TABLE table.
Hi Moritz,
I don't see any interesting messages related to this subject in the indexserver traces. My system is an AWS deployed development system without any formal SAP support, so I don't have any means of opening a message. Makes me wonder though if other development system users also experience these errors.
Regards, Dirk
Hello Martin,
are you able to offer any more insight into this problem? I'm facing the same problem as Dirk. Each time I stop my AWS HANA instance, the rollback table count increases. I've played around with the older statistics server, and the newer embedded stats server, but neither approaches clear/reduce the entries in this table. All of the Notes on this topic are fairly dead-ended (e.g. increase the thread worker heap size, stop/start HANA and the table should clear... well, the table doesn't clear, and we can't continually increase the thread stack size). I'm able to provide access to my system if you're willing to investigate. I imagine any users with HANA on AWS that is stopped/started via. the CAL console will eventually (are) experiencing this issue.
Look forward to your reply,
Hagen
Hi Dirk,
have you had any luck in resolving this issue? I've followed Niklas' advice, and boosted the number of statserver instances to 1, however my stat server instance sits at standby now. See below.
If I can get statserver Detail to switch to master, I expect the rollback table to be cleared. But how?!
I'm on Rev 91.
Cheers
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Dirk,
we had the same problem here, so we decided to upgrade our HANA version to the newest release on our AWS. After that, the table was still filled, sometimes also getting larger. There was only one error, I think in the index server trace, which told, there is a problem to connect to 127.0.0.1:30005.
I tried to activate the statserver via nameserver.ini, but it always got deactivated, the error message for the installation (nameserver trace) was just like "ERR:"
So I followed note 1950221... The important step in my case was step 7, the number of instances was still at 0 before. After fixing this, the statserver came up, the alerts where available again (another problem we had) and the rollback table got cleared.
Best regards,
Niklas
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
User | Count |
---|---|
95 | |
11 | |
11 | |
10 | |
9 | |
8 | |
6 | |
5 | |
4 | |
4 |
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.