Conference decwet::networker

416.0. "INCONSISTENCY and PERFORMANCE" by BACHUS::DUPLICY (Bruno DUPLICY) Mon Feb 17 1997 12:03

Hello,

I have a strange problem on a NSR V4.2B for Digital Unix 4.0b.

The nsr server runs on a Alpha server 4000 bi-processor.
The backup device is a TL812 (with 4 TZK88 an 1 jukebox)

The NSR backup server is configured with 4 drives, with 4 sessions on each
drive, so with a parallelism of 16.

The backup server should backup 15 clients with a total of 80 files systems each
night. So at a specific moment, it is possible that the server should backup
more than 16 sessions at the same time. (16 actives + some seesions in the queue
list)

For an external reason  ( CA unicenter installation ) last night the robot was
blocked. The result is that only the already loaded tapes where available for
the backup. The backups took more that twice the normal needed time to do the 
backup operation.

Question 1:
	In /nsr/logs/message.log  I have found the following message:
	nsrd: inconsistency in overall number of authorized save sessions,
		re-setting value from (4) to (5)
	server_backup: / saving to pool ....
        nsrd: inconsistency in overall number of authorized save sessions, 
               re-setting value from (5) to (6)
	server_backup: /usr saving to pool ....

	.... and so on up to (10) to (11).

	Is it possible that NSR is increasing by himself the value of maximum
	sessions authorized per tape drive?


Question 2:
	We are doing cloning based on SSID of the previous backup.
	The cloning normally works with a throuput of 1.2MB to 1.5MB /sec.
	But after the above situation ( I guess ) the throuput falls down to
	50KB to 200KB which is totally unacceptable to clone 50GB.

	Is it due to the increase of the multiplexing factor on each
	drives, and so the difficulties for NSR to retreive the datas for a 
	specific SSID?


Question 3:
	Is there a way to know what is the real mawimum seesion per drive used?


	Thanks in advance


	Bruno

	I don't know where the "re-setting value" messages come from
	but as to your question about maximum sessions per drive, that
	number is advisory.  NetWorker will try to keep the session count
	per drive to that level but it will exceed it if needed as
	long as the parallelism is not execed.  This can be affected
	by things like disabled drives and pool configurations.

	Since the software couldn't change tapes it may have tried to make
	do by moving saves from device to device as each tape filled.
	It's hard to say what all happened and it's probably not worth
	chasing since it only happened with a hardware problem.

	Please let us know if this happens without a known cause.