EVA4000 Slow Performance

[Fernando Boza]

I have a EVA4000 with a DiskGroupA which LUNs "SybaseProd1" and "SybaseProd2" whose propietary controller is the A.
In addition I have a DiskGroupB which a LUN "Respaldo" whose propietary controller is the B.
EVA4000 is connected by SAN to rp8400 with HP-UX v1. When the base of tados copies the data of the LUNs SybaseProd 1 and 2 to the LUN Respaldo, the system is put slow as much... that the commando bdf is delayed several minutes in responding.
I have been able to observe with the EVAPerf that during that backup to disc, the DiskGroupB has as much reading as writing... which seems to me strange since single it must have writing. I am in the correct thing or I am mistaken?

[JAguilar]

Have you looked at your controller logs on the EVA management console to see if your getting an port errors. I would also check the switch ports for any errors. If your getting any type of PVLinks errors on the RP box I found an article on the ITRC that fiber arrays will do this when configured w/the default timeout value in LVM which happens to be 30 seconds. When the machine is under heavy I/O this PVLink failed condition or a PVLink powerfailed condition can occur. The PVlink error is a system generated error that is reported when the fiber disk can't communicate w/the system w/in the allotted 30 second timeout period. You will then get a PVlink restored message w/in a few seconds. According to the article you can change this value w/the following command. pvchange -t 180 The 180 seconds is a value that EMC recommends on their Fiber arrays for HP-UX. These are some of the type of errors you might see:

Jan 12 10:36:37 fcprod vmunix: SCSI: isrEscape Controller at 1/0/6/0/0.
Jan 12 10:36:37 fcprod vmunix:
Jan 12 10:36:37 fcprod vmunix: SCSI: First party detected bus hang (HTH) -- lbol
t: 4177, dev: cb046002
Jan 12 10:36:37 fcprod vmunix: lbp->state: 30008
Jan 12 10:36:37 fcprod vmunix: lbp->offset: ffffffff
Jan 12 10:36:37 fcprod vmunix: lbp->nominalOffset: 450
Jan 12 10:36:37 fcprod vmunix: lbp->Cmdindex: 9
Jan 12 10:36:37 fcprod vmunix: lbp->last_nexus_index: 1
Jan 12 10:36:37 fcprod vmunix: lbp->nexus_index: 2
Jan 12 10:36:37 fcprod vmunix: uCmdSent: 9005a40 uNexus_offset: 2004560
8
Jan 12 10:36:37 fcprod vmunix: last lbp->puStatus [0000000041aa5600]:
Jan 12 10:36:37 fcprod vmunix: 00031600 00031600 00000000 00000000
Jan 12 10:36:37 fcprod vmunix: next lbp->puStatus [0000000041aa57f0]:
Jan 12 10:36:37 fcprod vmunix: 00000000 00000000 00000000 00000000
Jan 12 10:36:37 fcprod vmunix: From most recent interrupt:

Jan 12 11:12:52 fcprod vmunix: SCSI: Async write error -- dev: b 31 0x080500, er
rno: 126, resid: 32768,
Jan 12 11:12:52 fcprod vmunix: SCSI: Async write error -- dev: b 31 0x080500, er
rno: 126, resid: 65536,
Jan 12 11:12:52 fcprod above message repeats 5 times
Jan 12 11:12:52 fcprod vmunix: blkno: 2525408, sectno: 5050816, offset: 2586017
792, bcount: 32768.
Jan 12 11:12:52 fcprod vmunix: blkno: 2525376, sectno: 5050752, offset: 2585985
024, bcount: 32768.
Jan 12 11:12:52 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080500 Failed! T
he PV is not accessible.
Jan 12 11:12:52 fcprod vmunix:
Jan 12 11:14:29 fcprod above message repeats 16 times
Jan 12 11:14:29 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080100 Failed! T
he PV is not accessible.
Jan 12 11:12:52 fcprod vmunix: SCSI: Async write error -- dev: b 31 0x080500, er
rno: 126, resid: 32768,
Jan 12 11:14:29 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081400 Failed! T
he PV is not accessible.
Jan 12 11:14:29 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081700 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: Lost quorum.
Jan 12 11:15:14 fcprod vmunix: This may block configuration changes and I/Os. In
order to reestablish quorum at least 8 of the following PVs (represented by cur
rent link) must become available:
Jan 12 11:15:14 fcprod vmunix: <31 0x080100> <31 0x080200> <31 0x080300> <31 0x0
80400> <31 0x080500> <31 0x080600> <31 0x080700> <31 0x081000> <31 0x081100> <31
0x081300> <31 0x081400> <31 0x081500> <31 0x081600> <31 0x081700> <31 0x082000>
<31 0x082100>
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080200 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080300 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080400 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080600 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080700 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081000 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081100 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081300 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081500 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081600 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x082000 Failed! T
he PV is not accessible.
Jan 12 11:15:14 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x082100 Failed! T
he PV is not accessible.
Jan 12 11:18:06 fcprod vmunix: LVM: VG 64 0x020000: Reestablished quorum.
Jan 12 11:18:06 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080100 Recovered
.
Jan 12 11:18:06 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080200 Recovered
.
Jan 12 11:18:06 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080400 Recovered
.
Jan 12 11:18:06 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080500 Recovered
.
Jan 12 11:18:06 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x080700 Recovered
.
Jan 12 11:18:06 fcprod vmunix: LVM: VG 64 0x020000: PVLink 31 0x081000 Recovered
.

If you have multipathing capabilities check to see if all the ports to the EVA are active and have no issues.

Hope this helps,
Jose Aguilar-Systems Engineer
SunGard Availability Services,Inc.