- This topic has 36 replies, 15 voices, and was last updated 7 years, 9 months ago by
.
-
Topic
-
We converted to Cloverleaf 6 about 6 weeks ago. We are encountering a problem several times a day where a smat file is open in the gui, and clicking on any command button at the bottom, or the file menu option, causes the smat files to be renamed to
.old. . A message box pops up that says “The SMAT index file is inconsistent with messages.” It lists three possible causes, however none of these causes has occured. It gives you the Yes or No option of re-opening the SMAT file. The renaming has already occured, so you have no control over the renaming. No one else had the file open. It is likely that new messages were written to the SMAT file while I had the file open in the gui, but that should not be a cause for this behavior.
You must immediately go to the file system and rename the .old files to something else, because if it happens again to the same file, the previous .old files are overwritten and lost.
We have in our Site Preferences the SMAT History option disabled. We run the cycle save script which works fine.
Does anyone else have this problem and how did you overcome it?
Peter Heggie
-
Replies
-
-
-
-
-
-
This morning a set of smat files recycled to .old after one of our administrators performed a resend of messages. It appears that during this operation, or just after the operation, but before the smat file was closed, a new message came into the thread and was saved to the smat file. This may or may not have triggered the cycle save.
Also, I was asked to provide the settings in the Process Configuration gui. For this particular site, all processes do not have any functions selected or enabled, such as Automatic Smat Cycling.
I remember during the Inforum session in 2012 where enhancements and suggestions were discussed, mention was made about changing the accessibility of smat files, with attention focused on how multiple users or multiple updaters to the file were managed.
I have been told that in previous Cloverleaf versions, when a user has a smat file open and has edited a message, the updated message was saved in a .tmp file in the integrator/temp directory. I am not seeing this so far in a small test I just tried. Did this mode exist for 5.8.4? If so, has it changed for 6.0?
That would imply that the user has their own copy of the smat file in a temp file, and a smat file is never locked by a user and only Cloverleaf processes update the file. Is this how it worked?
One possible conclusion is that this new automatic cycle save is the result of trying to protect the integrity of the smat file from multiple updates, by carving off the existing smat file into a .old and writing new incoming messages to the new smat file. This is just a theory.
Can someone shed light on how multiple user access is managed?
Peter Heggie
-
-
yes we did hear back – we were asked to try setting the Process Configuration to enable Automatic Smat Cycling. I have enabled this today in our test environment for one of our processes; I should probably do it for all of them because we have not had a hit on the problem yet.
My guess is that the same underlying condition will still exist, that particular updates to a smat file while it is being used will cause an automatic cycle. With this option enabled, the cycle will result in the old smat file being written to the new smat sub-directory with a timestamp. This should resolve the lost data issue but may still be annoying in some ways.
I will wait until I see the behavior before posting more.
Peter Heggie
-
-
-
-
This issue first came about in 5.8.5 when there was an attempt to lock the smat file to a single user where before it was opened as read only to multiple uses. You could toggle and edit the message and resend it with the only evidence of the edit was in the /hci/cis5.8/integrator/temp directory in a file named with a random name beginning with tmp. It did not save the message you edited in smat as it does with this change. You never know when the cycle happens. You can be in it perform various tasks get out back in then out again and back in an hour later and all of a sudden it cycles with out notification/error or any mention of it you only notice that it has 0 records or that it has 1 record. Between the last time you got out and getting back in you do not know what happened to any and all messages between that time. If they are lost that is critical, there is no way of knowing.
-
-
-
-
-
-
-
-
-
-
-
-
-
-
Max:
I have seen some anomalies with SMAT when multiple people are viewing saved messages for a thread. Messages that should be there do not appear immediately in the SMAT. I don’t *think* we’re losing any messages, but it’s making me nervous.
Please confirm if this patch has been included in 6.1.1. We are about 4 weeks from a big go-live and don’t want to be having problems with SMAT data disappearing.
Jeff Dinsmore
Chesapeake Regional Healthcare -
-
-
-
-
-
I understand the WAL function of SQLite – and that all makes sense.
The issue I’m having is with SMAT files.
Let’s say I’m doing some testing. I send a message from system X bound for System Y and I’m saving outbound messages on the thread to “Y”.
I’m not always seeing a message in the SMAT file after I know it’s been sent to “Y”. It may show up a minute later, maybe right away – it’s not entirely predictable. There seems to be more frequent (apparent) write delays if more than one user is viewing data in a given SMAT.
This behavior is different than my experience with our current CL5.6 engines.
Our new engines are 6.1.1
Is CL 6.x not flushing the file handle(s) it’s has open to write to SMAT? That might give a nominal improvement in speed, but would reduce usability dramatically – and could also expose the SMAT files to data loss if there’s a power failure or if an engine process is killed forcibly.
Thanks!
Jeff Dinsmore
Chesapeake Regional Healthcare -
This started out as a discussion about SMAT files (traditional) getting cycled when other actions are being selected in the SMAT GUI.
It appears this has been identified as a bug and the correction to exist in 6.1 or 6.2.
But then the discussion morphed to SMATdb.
So the question still remains (at least in my mind) wass the original bug fixed and in what release?
email: jim.kosloskey@jim-kosloskey.com 29+ years Cloverleaf, 59 years IT - old fart.
-
-
-
We are seeing the random cycling in our new 6.1.2.0 (AIX7.1) test environment. It has happened to 7 smatdb files in the last 3 days. Seeing other weird behavior that seems to be related to this too. One of the smatdb files randomly cycled on Tuesday. The new file had messages in it for Wednesday and no messages for today(Thursday) even though the thread statistics showed it should have 22 messages. The file timestamp showed last activity from Wednesday. I turned up the engine logging and bounced the process and voila, all of today’s messages wrote to the smatdb with the correct time saved timestamp from when the messages actually passed through the thread.
-
-
- The forum ‘Cloverleaf’ is closed to new topics and replies.
