IBM ESS F20 on SAN causes extremely long boot times

[Jim Carter]

I've never seen this before and couldn't find it addressed in the forum archives, so here goes:

I've noticed when we have the Shark connected to the V2600s, zoned through the core and edge Brocades (FC16Bs), that the boot times go through the roof. Specifically, at the "INIT CEF2" phase, the systems spend 24 minutes before continuing. Before the Shark was connected (no other devices attached) the boot time on these servers was pretty normal.

Does anyone have any experience with this problem? I currently have only 2 LUNs definded on the Shark but will be defining 257 - I can't wait to see how that effects boot times!

[Duncan Edmonstone]

Jim, this is a bit of a wild stab in the dark, but here goes...

The slow boot *may* be connected with how HP-UX scans the IO buses during startup... traditionally HP-UX would look at every target on every controller, but would not look at every LUN if LUN 0 did not exist for this controller. This causes a problem with some disk arrays where LUN assignment does not necessarily start at LUN0 for every controller. To resolve this HP produced a patch to the SCSI subsystem which causes the bus scan to look at every LUN, but can significantly slow down boot times.

This has been addressed to some extent in patch PHKL_24187 - you might want to take a look at installing this patch.

HTH

Duncan

[Jim Carter]

Just in case anyone else pulls this from the archives, here is the solution:

The Brocade 3800 uses a different method to determine soft or hard zoning enforcement than the other Silkworms. The switch must be placed in soft zone mode to work successfully with the V2600s (and maybe others). To do so, you must use a combination of WWN and PORT numbers when you define zones (see the documentation CD that comes with the switches).

After I soft zoned the SAN, everything worked fine.