How to Control HANA Traces filling in File system?

[Former Member]

Hello Colleagues,

In the past, I noticed that in HANA Studio, the HANA traces size in Disk Usage was getting full for every 2 days. But now, it is getting full for every 10 minutes.

Our total trace volume size is 104, 96 (see attached screenshot) and we have no resources to increase it.

Are there any Options, to control this autofilling of trace files? From System Monitoring, I also see no users are logged on into the System in order to find the root cause.of the issue.

Any ideas, please?

Thanks & best regards,

Sreenu

[Former Member]

From the screenshot we can see that the actual trace size is small (1 GB) while the overall trace file system utilization is large (100 GB). So you have to check what else apart from SAP HANA traces occupies this file system unintentionally and eliminate the root cause. It doesn't seem to be a SAP HANA trace issue.

[davidebruno]

I'm agree with Martin

[Former Member]

Hi Martin,

Thank you for the Response.

But, all the logs that are getting full in log Directory are not in readable Format.

I assume that, the trace files getting full is due to the error "SAP DBTech JDBC: [3584]: distributed SQL error: [9] Memory allocation failed".

I get this error as soon as i remove the old log files and perform the new backup once the DB back in online mode.

Any ideas, please?

Best regards,

Sreenu

[Former Member]

Hi,

There are many threads which state your issue

Hana server management best practices | SCN

Also KBA

1997337 - "A communication error occurred, with the HDB TcpIp Server" 

Hopefully it helps

Br, Vaibhav

[Former Member]

The log area contains redo logs and not traces. Redo logs are not readable via a text editor, they contain change information. You have to schedule log backups on a regular basis.

[lucas_oliveira]

Hello Sreenu,

First of, if the trace mount is getting filled too fast, verify if there are traces active on your system. Often we see customer with sql trace activated all the time for no good reason. Other component tracing will surely affect that. If there are unnecessary traces activated, just deactivate them and clear the traces.

There's a nice sql query ready on SAP Note 1969700 that can help you with that (HANA_Traces_ActivatedAndModifiedTraces_Rev70+).

Now regarding maintenance: you need to make sure file rotation is on (see Configure Trace File Rotation - SAP HANA Administration Guide - SAP Library ) and use a linux cron job to perform trace archiving. No magic

BRs,

Lucas de Oliveira

[Former Member]

Hello Lucas,

Thank you for your Response.

"HANA_Traces_ActivatedAndModifiedtraces_Rev70+" SQL query showed no results.  So, there are no active traces.

If I understood correctly, do i have to decrease the Default value 10000000 of maxfilesize parameter?

Best regards,

Sreenu

[lucas_oliveira]

Well, 10mb is the default and I think it's an ok value for file rotation so you may leave that as it is and check why your trace files are getting big so fast.

Chances are that you have an error trace somewhere (indexserver / nameserver /etc) dumping loads of info and that could be the root cause of this scenario. Just get that cleared.

Another possibility is the current trace files you got. Any huge file in there? If so, move/archive/delete it.

BRs,
Lucas de Oliveira

[Former Member]

Hello Licas,

Actually, all the logs that are getting full are from log Directory.

I am able to figure out the error trace because all the logs that are getting full in log Directory are not in readable Format.

I am deleting them manually bu the question is, how to control these trace logs? I can't delete them for every 10 minutes.

Any ideas?

Thank you,

Sreenu

[davidebruno]

Hi,

sorry but what it the content of the traces directory?

Is it possible that in the same file system, there is the directory for the backup log?

Please check under /usr/sap/<SID>/HDB<SysNum>/backup/log

[Former Member]

Hi Davide,

Yeah, I checked under the path /usr/sap/<SID>/HDB/<SysNum>/backup/log but I don't see any Information in order to troubleshoot the issue.

Thank you,

Sreenu