Exalead and slow disks

Build 1501 on 14/Nov/2017  This topic last edited on: 25/Aug/2014, at 13:08

It may happen that you cannot find some items in the database, although you are sure they exist.

This may be due to the unprocessed records in the indexQueueTable.

Check the number of records in the indexQueueTable. If it holds many thousands records it is possible that Back4 was not able to push them in chunk to Exalead.

We started from a clean situation by cleaning the indexQueueTable and we could see that the new objects are pushed to the Exalead index.

Monitoring the server we noticed that the Disk Queue Length on the Exalead server is always greater than 1 and the Active time is almost 100%
 
exaleaddisk1

This is telling us that the disk is not fast enough and the information are queuing. This cause Exalead to freeze from time to time.

This can also be see in the Back4 log:
 
exaleaddisk2
 
where from time to time the commit operation timeouts.

In a long run this might cause the indexQueueTable to be filled up again. For the time being it seems to keep up.

The records in the indexQueueTable are increasing and then decreasing but it's very close to its limit. If for some reason the Back4 is stopped for a while, it will probably not be able to recover.

This is due to a hardware limitation, in other words, the too slow disks