Hello Neil, VxVM gives you no errors if there are no problems in SAN.
SAN can be a problem even if there are seen no critical errors on switches. Cables , buffer credits , 8Gbps Brocade Fill Word etc. If hba is ok, if frond-end VMAX port is ok, if cables are ok (and switches) then VxVM is ok as well Regards, Pavel -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Neil Swallow Sent: Thursday, November 29, 2012 4:54 PM To: [email protected] Subject: [Veritas-vx] EMC VMAX with VxVM 5.0 Hello all, Just after some feedback on whether anyone has experienced anything similar to an issue we're currently seeing. We're trying to migrate from an HDS SAN to an EMC VMAX on a Solaris 10 server. We have the server up to supported levels for drivers etc. and have removed one of two paths to the HDS and routed that to the EMC and all is seen OK. However when we try using the EMC (mirroring the volumes from the HDS to the VMAX) we are inundated with scsi transport errors and a lot of i/o wait time. DMP (VX, not powerpath) is disabling the path and re-enabling it at random points (not necessarily during mirroring) and even if the mirror completes the plexes on the VMAX LUN are disabling eventually and vxdisk list reports the LUN as failing. I've manually disabled the path and created a separate (vxfs) fileystem on the LUN and done some file copies etc and still get problems so I'm not convinced it's a DMP issue. I've also done the same with a UFS filesytem and oddly UFS provides the least amount of i/o waiting and out-and-out read/write failures of the copy commands etc. but still registers lots of transport errors. With no issues on HDS I'm loathe to blame any of the VX elements but EMC are providing no help since the VMAX itself and all the switches are reporting no issues and they are trying to blame vxvm/vxdmp/vxfs. Whilst I agree that things appear better without any VX involved, the issues do not go away entirely so I am currently putting that down to a lower tolerance to issues on the part of the various VX elements. Anyway, if anyone has any idea/suggestions/similar issues (any additional logging I can add at a VX level to help prove anything) that have been rectified, I'd appreciate it thanks, Neil _______________________________________________ Veritas-vx maillist - [email protected] http://mailman.eng.auburn.edu/mailman/listinfo/veritas-vx _______________________________________________ Veritas-vx maillist - [email protected] http://mailman.eng.auburn.edu/mailman/listinfo/veritas-vx
