06-17-2011 01:42 AM
I have mixed 48000 and DCX environment. Yesterday the UNIX team did a emergency (but planned) power off/on on one Power 780 system and did change power supplies. After that some of the LPARS began loosing disks. In particular one of the LPARS we still cannot see the disks previously assigned. It is zoned with a midrange disk (HDS 9585V), all other LPARS connected to enterprise disks (HDS USP, HDS USP-V and IBM DS8000).
I see all director ports as F-port, Online, Sync but still no disks. I looked for something odd and saw a high Rx uWatts values on director ports from that AIX box. On 4 of them I saw 508, 570, 568 and 607 uWatts.
I would want to know
- what is the nominal / max RX uWatts power ?
- can it be the reason for not seeing the disks ?
- is it correctable and how ?
06-18-2011 02:53 AM
the max and min power values are sfp specific. You have to check the tech specs for each SFP type on the vendor web site.
No it can not be the reason if the you have no increasing error rates on the ports.
Did you have checked your error counters after the reboot?
Before you check them please clear all counters with the commad statsclear. Wait 5 to 10 mintes and check the error counter with porterrshow
Did you have WWN based zoning or port based zoning inplace?
If the second one is the case I would assume that some connect FC cables into the wrong HBA ports. This can be the reason why the server can not see the disks.
Can you check this?
I hope this helps, if not please let me know,
06-19-2011 08:51 AM
Brocade best practice for shortwave SFPs is between 110 and 490 uW although I've seen up to 580 uW launch power for 8G modules which I believe is normal.
Longwave SFPs are different. These can easily have up to 2 mW launch power especially for long range types (CWDM/DWDM SFPs).
SFP vendors such as Finisar or JDSU list Tx power and Rx sensitivity limits in their specification sheets.
No light would be a clear indication of link failure. Poor signal alone can't be responsible for high-level problems such as missing targets. Normally there are other indicators on lower levels of the FC stack (such as CRC errors, sync loss incidents and so on) which in turn cause login issues or communication failures that translate to FCP-level problems.
Hope this helps,