Conference 7.286::fddi

2063.0. "MSCP command timeout on FDDI MSCP served disks" by GIDDAY::STANISLAUS () Wed Jun 12 1996 08:33

    Also posted in Cluster, Gigaswitch, VMSnotes conferences.
    
    
		MSCP COMMAND TIMEOUT IN THE FOLLOWING CONFIG:
		---------------------------------------------

	There are no FXA0 or PEA0 or any other errors on the cluster nodes,
except for repeated MSCP COMMAND TIMEOUT errors in the errorlog. One node
repeatedly says Mount Verification in progress on a shadow set DSAn and 
immediately followed by Mount Verification completed message. About 80 secs 
later another Mount Verification in progress on the same DSAn or another DSAn
and again immediatley followed by Mount Verification completed. 

	There are no Gigaswitch errors seen with a MIB browser.

SDA>SHO LAN/FULL/DEV=FXA on all 3 nodes see no Last Fatal Error or Previos
Fatal Error.

	Above only happens during extreme disk activity and as well as backup
is taking place - mostly at night. During day no problems.

	Config:

	Site 1 has a VAX 7620 (ASX250) and another VAX 7620 (ASX251) connected
via CI (CIXCD) and via FDDI (DEMFA) using a Gigaswitch. There are several disks
connected via HSJ40 controllers.

	Site 2 has a VAX 7620 (ASX150) connected via CI (CIXCD) and FDDI (DEMFA)
using a Gigaswitch. There are several disks connected via HSJ40 controllers.

	Most shadow sets are 3 member sets with 2 members on site 1 and the
3rd member at site 2.

	Sites 1 and 2 are interconnected via the above Gigaswitches AGL cards
and Microwave equipment running at 45 Mbs. This is a BRS (business recovery
type of configuration).

	Problem:

	Everything was working fine for the last year or so. Last week more
SCSI disks were added and the Gigaswitch F/W was upgraded from 2.2 to 3.x
where x=0 or 0.1 (I am not sure). The customer also said the load on ASX250
is now about 80% compared to 50 or 60 percent before last week.

	From last week ASX250 reports the above mount verification in progress
and mount verification completed messages repeatedly during heavy disk write
activity. The errorlog says MSCP COMMAND TIMEOUT pointing to ASX150 MSCP server.
No other errors AT ALL on any member of the cluster. The errors stop if the 
customer dismounts member 3 across the microwave, so that the shadow set becomes
a two member local shadow set.

	Versions:

	VMS 5.5-2 on all 3 nodes
	FXDRIVER X-22A1 (link date 4th Mar 1994) on all 3 nodes
	SHDRIVER X-89A1A4 (link date 30th Mar 1995) on all 3 nodes
	DSDRIVER X-19A19A1A3 on all 3 nodes
	PEDRIVER VMS-5 on ASX250 and ASX251 at site 1 and VMS-1 on ASX150 at 
	site 2.
	DEMFA F/W at 2.0 on all 3 nodes.
	Both Gigaswitches upgraded from 2.2 to 3.x F/W last week.


Customer doesn't want to revert to Gigaswith 2.2 if there is no known problem 
with version 3.x.

	Could it be that PEDRIVER on ASX150 is VMS-1 image and not VMS-5
like the other nodes. ASX150 is MSCP serving the shadowed disk to AXP250
and AXP251 across the DEMFA/FDDI/Gigaswitch/Microwave. ASX251 is not as
heavily loaded as ASX250 and so that may be the reason why ASX251 is not 
reporting the about mount verification messages.

	Any suggestions apart from patching ASX150's PEDRIVER to VMS-5
like the other 2 nodes.

Alphonse