Missing Record During Sync Process

[bradohare]

Hello,

We recently had some odd behavoir during a sync between two instances. Basically, a sync process ran to move 4 records and only 3 came across. We are on 9.5.1 and opened a ticket with IBM about it as well. I was just curious to see if anyone else had experienced something similiar. Here are the details:

1. We have two instances “input” and “reporting”. Our “reporting” instance syncs with the input many times throughout the day. This has been running for awhile in a PROD capacity.

2. The issue specifically refers to a single sync in which 4 records where loaded (all other syncs were fine during the day) and only 3 were reflected in the synched cube

3. There process that loads data in the “input” instance and the following validations were done on this instance:
--------------Does an AsciiOutput for each record that loads and we were able to see the 4 records
--------------We reviewed the Transaction Log on the “input” instance via the GUI and saw the 4 records
--------------We reviewed via the cube Viewer and the data was reflected as anticipated.

4. The “reporting” instance than synched with the “input” cube
--------------We’ve then reviewed the “TM1ReplicationLog_XXXXXXXXXXXXXX_input.log” for the appropriate time frame and saw the 4 four records reflected as anticipated
--------------We then reviewed transaction log on “reporting” instance via the GUI and saw the 4 records as anticipated
--------------When we went to the cube viewer and only three of the four were displayed
--------------When then verified throughout the rest of the transaction log that there were no other entries that would have unintentionally altered the data for this intersection

5. This single record was missing in the cube viewer throughout the week. However when we had our weekly reboot, upon the server coming back up all 4 of the records now properly were reflected in the cube viewer.

It almost seems like for some reason when the sync executed the transaction moved however wasn’t updated in memory. Then after being rebooted, the server re-reads the transaction log and populated the data as anticipated.

Any thoughts?

thanks
brad

[mattgoff]

My only thought is you have a 100% chance of your PMR being closed as "cannot reproduce." Sorry.

Matt

[bradohare]

Matt, I agree with your assement. Opening a case is more of an exercise in futility.

[Steve Vincent]

were those data records or attribute values? i've recently given up with replication and wrote my own, i couldn't handle the random nature of what it did/didn't do anymore

[bradohare]

Hi Steve,

They were data records. When you said you wrote your own replication, from a technical perspective could you provide a little insight into how/what exactly you mean? We've been running in a high volume production environment for about a year now. This is the first time we've experienced this behavoir. Obviously, once is one time to many, but still.....

thanks
brad