hcicmd save_cycle issue

[A Lim]

Hi,

Has anyone ever encounter intermittent issues when the auto cycle save script using the hcicmd save_cycle command that it did not managed to cycle properly, thus no messages were saved in the folder?

While I understand that if the thread is down, it will not save; but this thread is running daily, and this issue only occurs intermittently, and has no error messages or any information recorded into the log.

Would appreciate if someone could give some idea on scenarios where save_cycle would not be able to work properly.

Thanks!

Topic

[Russ Ross]

One scenario where I’ve had problems like you described are when I accidently configured 2 interfaces in the same site to use the same SMAT file.

This can happen when you copy a thread to make a new one and over look changing the SMAT file name.

I think I also had SMAT file cycle issues when the SMAT file got extermely large or reached our system limit on file size of 2 gigabytes.

These days we cycle our SMAT files ever 3 hours becuase our message flow is too high at about 6 million messages a day to only cycle SMAT once a day.

Russ Ross
RussRoss318@gmail.com

[A Lim]

Russ,

Our daily filesize is not really that large, have not even reached 10MB yet, so I don’t think it’s a filesize limit issue.

I have check the SMAT filenames and they are definitely difference as they are using it to do some reconciliation matching (as a double checking system).

The only possibility that I could think of is that at the present moment when the save_cycle was run, there were things running in the thread, thus making it not possible for save_cycle to run. But I don’t exactly have any documentation or information on what could cause this to fail.

I always thought this would be part of a FAQ Q&A… But it does not look like anyone else has such issues before.

[Russ Ross]

Let us consider this possibility.

If your cycle_save script is similar to ours it goes something like this:

– cycle SMAT to create something with “old” in the file name

– then archive the “old”/cycled SMAT file

If that is your case you could run into a problem if someone manually cycles SMAT so an “old”/cycled SMAT file already exists and gets blown away when your cycle SMAT script runs, making it look like SMAT lost track of messages.

The way I fixed this problem was to archive any “old”/cycled SMAT files before cycling SMAT and then archive the “old”/cycled SMAT again.

This still does not prevent a problem when someone manaully cycles SMAT twice before the “old”/cycled SMAT file has a chance to be archive before being overwritten.

If someone cycled saved SMAT twice in a row quickly with your slower message flow that might create an empy file.

This might occur if someone meant to cycle SMAT “in” and the SMAT “out” but ended up doing the same one either “in” or “out” twice.

You might have to put some debug in your cycle save script to help gain more insight.

For example some debug output to show the site/thread information being cycled and a directory listing before and after each action takend of the process directory.

Also pay attention if the same thread is having the problem or does the problem seem random.

Russ Ross
RussRoss318@gmail.com

[Charlie Bursell]

Are you logging the output of the hcicmd command?

Could simply be that the process is too busy to process the command when issued.  Look for response like “No response within timeout”

[A Lim]

Well, according to the log, the save cycle was successful, but there is no regular msg files found. The next day msg files does not contain the previous day ones that went missing, so looks like the save cycle did happen, but something happened somewhere and causes the old.msg files to be missing.

The script was not written by me, I inherited the site, and this is the 1st time I am looking at the script! I do know that the script remove the “old.msg” ones, then do a save cycle, and then rename the .msg files to something else instead of “old.msg” like “date/time.msg”. I am not sure if it was the save cycle that causes it or something else – like someone did a manual save cycle before it triggers.

But there were no error messages in the auto save cycle logs, and the old messages was not retained from the thread in the next save cycle time frame.

[Author]

Replies