ES40 Redundant Supplies - Unexpected System Restart

[Robert R. Hickman]

We have a pair of rack mounted Model 2 ES40's configured in a TruCluster. Each machine has 2 CPUs, 4GB of RAM, 3 Power Supplies, and all disks are in external shelves (with their own power supplies). According to the ES40 QuicSpec, each system only needs one power supply to run, and that seems to be true. We have dual power feeds, with supply "0" on feed "A", and supplies "1" and "2" on feed "B". If we de-energize feed "A" (supply "0") the system just keeps on working (from supplies "1" &"2"). However, in a recent "experiment", de-energizing feed "B" (supplies "1" and "2") caused an unexpected restart attempt by SRM and/or Tru64; i.e., with one supply ("0") still active, the system should have kept going, but instead tried to restart itself and ended up (because of autoaction = halt?) at the SRM prompt. At that point the system could be manually booted and would run on one supply ("0"). Does anyone know why it tried to restart on the loss of two (of three) supplies? Is this an SRM issue, a Tru64 issue, or both?

Bob Hickman, National Grid USA

ps - system is Version 5.1, with no patches applied.

[Dave Bechtold]

Hi Robert,

The restriction for the 1 Power Supply system is as follows (Model 2)

Maximum configuration with one power supply:
- Two CPUs
- One storage cage (six hard drives)
- Four to twelve DIMMs

I'm not aware of any software issues which would account for the symptom you reported.

Maybe Power Supply 0 is marginal and needs to be replaced. Maybe rotate it with one of the other Power Supplies.

Hope that helps,
Dave Bechtold

[Robert R. Hickman]

Dave,

Your point about the DIMM count appears to be the jackpot. The original ES40 User Guide didn't say a thing about DIMM count - just the number of CPUs (we have 2) and occupied drive cages (we have none). I missed (in the latest ES40 quickspec) the DIMM count limitation. We have 4 GB of memory, in the form of four 1 GB memory features (a total of 16 DIMMs). Thus we're over the 12 DIMM limit mentioned in the quickspec, and will have to remove some memory if we want to safely operate on only one of two independent power feeds (each with its own UPS); i.e., complete redundancy is more important to us than performance. I wish the ES40's maximum power supply count wasn't an odd number (3); it creates a headache when trying to deal with dual redundant power feeds.

Thanks again for your answer.

Bob Hickman

PS - Between the time I posted the question and the time of your response, my ITC user ID changed (because my initial registration was (by accident) on the European site). Thus I'm not able to "assign points"; I consider your answer a "10".