Every time CI is coming into UCMDB it compared vs content of UCMDB and vs data coming in the same bulk. The operation repeated for every CI.
The default for the fuse is 100K. It's highly unrecommended to raise the number. You'll have significant performance problems and possible Out of Memory situation at the end.
Instead of it, the Discovery Job used for data population need to be changed to sent data in smaller chunks.
In case you see this in OTTB job please let us know the name of the job. We could have ready solution, or new problem will be created.
I suspect this might be coming from the specific job Host Connection by WMI, running against specific IP.
Could you try to run the job excluding this IP and confirm once ?
If you can confirm it is happening with specific IP, then needs to understand why only this specific ip , getting so much data or data with less attribtue values, which could be probable cause of the problem.
I dont think it was from a specific job, because I was seeing the same error when runging Host Connection by SNMP.
I follow Dima and lower the chunk size of result down to 5000. Is this a good number to be at? and is this the change you were talking about Dima?
# probe mgr - the max size of task results chunk to send. if large chunk of results received,
# it will be break into smaller ones in order to prevent overload on the server
appilog.agent.local.maxTaskResultSize = 5000
Lowering of the chunk size could be a good idea.
I got an answer from R&D regarding following fuses:
The first one is the maximum amount of CIs that we can handle internally (due to merges, identification, etc…).
The second one is the maximum amount of CIs that we allow to accept as the Data-In bulk. Afterwards, the Data-In will fetch more CIs from the UCMDB for the analysis of the input-CIs. Or in another words – the second one is the amount of CIs/relations in the input and the first one is the total amount of CIs (including the ones that were needed for the analysis of the input).
is there any optimal number for the chunk and the two fuses? What is the optimal combination and what are the factors that these numbers depend on? Only the number of CIs that the TQL returns?
How could you avoid the identification map growing? Is transferring all attributes of the CIs a good or a bad idea?