[omniORB] Omninames crashing during Checkpoint operation

[Jayaraman, Thirupurasundari]

All

We are observing a crash in omniNames randomly and we get the following log whenever the crash happens.

We are using version 4.0.7 and 4.1.4. We are seeing this issue in both versions.

We are running omni on Linux, Open SUSE 11.2 OS.

Mon Jul 16 14:02:06 2012:

Starting omniNames for the first time.
Wrote initial log file.
Read log file successfully.
Root context is IOR:010000002b00000049444c3a6f6d672e6f72672f436f734e616d696e672f4e616d696e67436f6e746578744578743a312e30000001000000000000007c000000010102000b00000031302e33392e32302e340000fbbe00000b0000004e616d6553657276696365000300000000000000080000000100000000545441010000001c000000010000000100010001000000010001050901010001000000090101000300000014000000010000000a0000003132372e302e302e3200f90a
Checkpointing Phase 1: Prepare.
Checkpointing Phase 2: Commit.
Checkpointing completed.

Mon Jul 16 14:17:06 2012:

Checkpointing Phase 1: Prepare.
Checkpointing Phase 2: Commit.
Checkpointing completed.

Tue Jul 17 12:02:06 2012:

Checkpointing Phase 1: Prepare.
Checkpointing Phase 2: Commit.
Error: failed to link backup file '/tmp/log_dir/omninames-DBBAMQ2.bak' to old log file '/tmp/log_dir/omninames-DBBAMQ2.log'.

When we checked the code, we found it is related to "checkpoint()" function in log.cc file.

http://fossies.org/dox/omniORB-4.1.6/log_8cc_source.html


We need to get this bug fixed soon because it may potentially impact our customer base once they start using our application heavily.

Thanks for your help,

Regards
Sundari.

[Agarwal, Shelendra]

You are starting omniNames with -logdir as /tmp/log_dir/
and /tmp/log_dir/omninames-DBBAMQ2.log is missing because /tmp/ is cleaned.

Try to start the omniNames with -logdir to a location where nobody deletes the log file.

Thanks & Regards,

Shelendra Agarwal

-----Original Message-----
From: Jayaraman, Thirupurasundari [mailto:Thirupurasundari.Jayaraman at kla-tencor.com]
Sent: Tuesday, August 14, 2012 6:26 PM
To: omniorb-list at omniorb-support.com
Cc: DL-Win-DBB-KTI
Subject: [omniORB] Omninames crashing during Checkpoint operation

All

We are observing a crash in omniNames randomly and we get the following log whenever the crash happens.

We are using version 4.0.7 and 4.1.4. We are seeing this issue in both versions.

We are running omni on Linux, Open SUSE 11.2 OS.

Mon Jul 16 14:02:06 2012:

Starting omniNames for the first time.
Wrote initial log file.
Read log file successfully.
Root context is IOR:010000002b00000049444c3a6f6d672e6f72672f436f734e616d696e672f4e616d696e67436f6e746578744578743a312e30000001000000000000007c000000010102000b00000031302e33392e32302e340000fbbe00000b0000004e616d6553657276696365000300000000000000080000000100000000545441010000001c000000010000000100010001000000010001050901010001000000090101000300000014000000010000000a0000003132372e302e302e3200f90a
Checkpointing Phase 1: Prepare.
Checkpointing Phase 2: Commit.
Checkpointing completed.

Mon Jul 16 14:17:06 2012:

Checkpointing Phase 1: Prepare.
Checkpointing Phase 2: Commit.
Checkpointing completed.

Tue Jul 17 12:02:06 2012:

Checkpointing Phase 1: Prepare.
Checkpointing Phase 2: Commit.
Error: failed to link backup file '/tmp/log_dir/omninames-DBBAMQ2.bak' to old log file '/tmp/log_dir/omninames-DBBAMQ2.log'.

When we checked the code, we found it is related to "checkpoint()" function in log.cc file.

http://fossies.org/dox/omniORB-4.1.6/log_8cc_source.html


We need to get this bug fixed soon because it may potentially impact our customer base once they start using our application heavily.

Thanks for your help,

Regards
Sundari.