SAP HANA log backup frequency

[Former Member]

Dear all,

we configured hana to use the hdbbackint backup interface.

Now we checked and have seen that over 650 logs are saved to the backup library one a day.

logs are genereated randomly but at least every 1 - 3 min.

Parameters are set :

log_backup_timeout_s = 900

log_segement_size_mb = 8 

i attached a screenshot of the log backup catalog.

do you know how we can handle that ? backups every 15 min should be ok i think.

greetings

oliver

[michael_gerstmayer · Accepted solution]

Hil Oliver,

Which service generates the high amount of log-segments? You can use the below SQL Statement to aggregate the entries of the backup catalog or the backup.log to check that:

service_type_name, count(*) from "PUBLIC"."M_BACKUP_CATALOG_FILES" a join "PUBLIC"."M_BACKUP_CATALOG" b on a.BACKUP_ID = b.BACKUP_ID where b.ENTRY_TYPE_NAME = 'log backup' and source_type_name = 'volume' and state_name = 'successful' group by service_type_name

If it is the Statisticsserver, you can e.g. think about migrating to the embedded statisticsserver service (SAP Note 1917938) or increasing the log_segment_size_mb to 1024 Mb accordingly (which is the default for the indexserver). I think 8 Mb is the default log_segment_size_mb for the XSEngine which could be increased to 1024Mb too. It is recommended to stay with the default setting of log_backup_timeout_s (time until a not yet full log-segment is closed) as it has an impact on the Recovery Point Objective as outlined in the SAP HANA Administration Guide.

Best regards,

Michael

[Former Member · Accepted solution]

Hi Michael,

the output is :

;COUNT(*)

1;3.011  

2;1.174  

3;2.994  

4;2.225  

the statistic server is already embedded.

why should i increase the log_segment_size_mb ? the logs are already switching before they reach the 8 MB.

greetings

Oliver

[michael_gerstmayer · Accepted solution]

Hello Oliver,

Unfortunately when pasting the SQL statment into the reply some columns got lost.. however, I was wondering if you can create an incident on component HAN-DB-BAC and provide a HANA Studio service connection to the system affected? At the moment it is not clear to my why log backups are way smaller than the configured size (I expect to see that only when the log_backup_timeout_s was reached) and written every 1 to 3 minutes (which is much less than the timeout).

Best regards,

Michael

HANA Product Support

[Former Member · Accepted solution]

i opened a ticket, i will post the solution when i got one.

[michael_gerstmayer · Accepted solution]

Hello,

it turned out that - on grouping the entries in backup.log by the service and filtering by states LogBackupStarted/LogBackupSuccess - a log backup is performed every 15 minutes. This is due to the log_backup_timeout_s which is reached as the log-segment was not full after this threshold exceeded.

2015-01-15T14:45:07+01:00  P038871      14aedd63d50 INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogStarted
2015-01-15T14:46:20+01:00  P038871      14aedd63d50 INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogSuccess

>> 15 Minutes

2015-01-15T15:00:07+01:00  P038871      14aede3f8ed INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogStarted
2015-01-15T15:01:20+01:00  P038871      14aede3f8ed INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogSuccess

>> 15 Minutes

2015-01-15T15:15:07+01:00  P038871      14aedf1b48e INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogStarted
2015-01-15T15:16:20+01:00  P038871      14aedf1b48e INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogSuccess

>> 15 Minutes

2015-01-15T15:30:07+01:00  P038871      14aedff702a INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogStarted
2015-01-15T15:31:20+01:00  P038871      14aedff702a INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogSuccess

Without grouping according to the service, it looks like a log backup happens every couple of minutes:

2015-01-15T14:29:04+01:00  P038871      14aedc754db INFO    LOGBCKUP state of service: nameserver, linux:31001, volume: 1, BackupLogSuccess
2015-01-15T14:31:20+01:00  P038871      14aedc881b0 INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogSuccess
2015-01-15T14:39:30+01:00  P038871      14aedd0e2c7 INFO    LOGBCKUP state of service: xsengine, linux:31007, volume: 2, BackupLogSuccess
2015-01-15T14:44:04+01:00  P038871      14aedd5107c INFO    LOGBCKUP state of service: nameserver, linux:31001, volume: 1, BackupLogSuccess
2015-01-15T14:46:20+01:00  P038871      14aedd63d50 INFO    LOGBCKUP state of service: indexserver, linux:31003, volume: 3, BackupLogSuccess
2015-01-15T14:54:21+01:00  P038871      14aedde7c02 INFO    LOGBCKUP state of service: xsengine, linux:31007, volume: 2, BackupLogSuccess

Best regards,

Michael

[Former Member]

Log backup will happen for every 15 mins (log_backup_timeout_s=900) only if your HANA is idel. however, th frequency depends on the activity performed on the system

Regards

Pavan Gunda

[Former Member]

Hi Olivier,

Although the parameter below is defined on the HANA system

log_backup_timeout_s=900

The backups occur very frequently depending on the changes written in the log segment

You will observe the following message in the BACKUP CONFIGURATION TAB when you point to *i* icon beside the BACKUP INTERVAL

*The backup interval you specify is the longest amount of time that can pass before a new log backup. Log backups may be written more frequently if there is more Database activity*

You must see log backups occurring at lower frequency when HANA is idle

Most importantly each service has its own log backup so the parameter is allowing us to set the time out when Database is idle, when HANA is active it will write log backups more frequently and we may not control that