IPP MHPCC Production Cluster Status

(Up to IPP for PS1)

Log of Work Done on Production Nodes

List of Spares for Production Nodes

Status as of 2011-01-20 ipp005,006,007,008,009,0012,0014,0016,0025 all need either USB or new cd/dvd drives with compatible cables to house the boot disk.

Wave1 Refurb Stage3 (Motherboard upgrades on ipp005,006,007 and 025.)

*ipp005 resulted in 2 degraded drives - rebuilding. *ipp006 CPU1 Memory slot closest to CPU not working. memory moved to black slot - all 24 GB working *ipp0025 CPU0 Memory slot furthest from CPU not working. memory moved to black slot - all 24 GB working

Wave1 Refurb Stage2 (Disk upgrades on ipp008,016, and 021, and motherboard

upgrades on ipp008 and ipp016.) ===

• no problems

Wave1 Refurb Stage1 (Disk and Motherboard upgrades IPP009,012,014)

components upgraded: motherboard + cpu + memory + 2TB HDD

• ipp009 : BIOS detects 6 memory modules in temp readings, BIOS/OS only recognizes 20GB of 24GB
• ipp012 : unexpectedly crashes under high rsync load - RAID card replaced.
• ipp014 : erratic beeping, suspected PSU, seems to be originating from PSB controller for HDD and fan display alarms, not mobo

Storage upgrade

• ipp005 : 2TB HDD install - Mohr (2010-07-02)
• ipp006 : 2TB HDD install - Bill/Cindy (2010-08-06)
• ipp007 : 2TB HDD install - Bill/Cindy (2010-08-06)

MHPCC System Maintenance

• Required to shutdown Pan-STARRS IPP (for 1hr starting @ 10AM HST) cluster in order to install power meters on IPP circuits.

Nodes down

• ipp008 : motherboard problems -- CPUs were changed on 2010.02.17, but still no boot
• ipp018 : motherboard problems? power supply problems?
• ipp037 : long history of return, etc.

Under-Used Nodes : Status 2012.02.09

• ipp004 : currently used for addstar / dvo databases
• ipp005 : recently replaced CPU from ipp018, has been used for processing since 2/16, but not yet for nebulous
• ipp008 : dead hardware (CPU appears to be defective 2010-02-22), (2010-03-22 Bill G. replaced motherboard, using CPU fans from IPP018)
• ipp009 : had several crashes in 2009, only used for nebulous since ~2009.10.29
• ipp012 : several crashes, suspected faulty memory, 8GB memory replaced (2010-08-06 Bill/Cindy)
• ipp014 : had several crashes in 2009, only 'available' in nebulous, not 'allocated' (since ~2009.10.29), suspect bad memory (mcelog entries ~2008)
• ipp016 : some crashes in 2009, only nebulous for some time (memory replaced 2010-05-28)
• ipp018 : dead hardware (motherboard dead; CPU's installed into ipp005 2010-02-16); replaced motherboard/cpu (2010-04-01)
• ipp025 : paul raised suspicions about memory problems; only used for nebulous since then. (memory replaced 2010-05-28)
• ipp027 : high loads, sluggish, suspect bad cpu, possible bad memory (mcelog entries dated daily) - 2012-02-09 Haydn flashed BIOS -> v1.04
• ipp037 : dead hardware (determined motherboard & CPU are defective 2010-02-22)
• ipp049 : seems to have somewhat slow RAID I/O; has not been used for processing, and only 'available' in nebulous
• ippc18 : battery temperature warning (Cindy added fans with the intention to cool RAID controller & batt., no warnings since 2010-02-22)
• ippc19 : battery temperature warning (Cindy added fans with the intention to cool RAID controller & batt., no warnings since 2010-02-09)
• ippdb02: series of crashes, motherboard swapped. ippc17 pstamp/datastore moved to ippdb02 2010-02-22.

Other suspicious Wave 1 machines

• ipp006 : a few crashes in 2009
• ipp010 : several crashes, kernel panics in 2009
• ipp017 : several high-load, high-memory crashes 2009-2010 (not a Wave 1 machine!)

Notes

• ipp005 - The memory was completely swapped and crashes continued. CPUs were replaced (2010.02.09). In production since 2010.02.16.
• ipp014 - using the chassis from ipp017 with disks from ipp014 (chassis had been ipp037, but was sent back to ASA for repair, ASA shipped back with replacement motherboard (Tyan S2912G2NR-E) Tyan S2912G2NR is EOL - kernel 2.6.31.5 is required.)
• ipp017 - using the chassis from ipp037 with disks from ipp017
• ipp018 - replaced 8GB of memory (2009-12-30)
• ipp027 - original areca card replaced with a newer 3ware card; using this raid as manoa backup
• ipp034 - using IPP037 Areca RAID Controller (IPP034 original Areca Controller appears to be defected -> Controller#1(PCI) H/W MONITOR DRAM 1-Bit ECC)
• ipp037 - using the chassis from ipp014 with disks from ipp037

• ippdb00 - in production as nebulous server
• ippdb02 - memtest86 4 passes completed no errors. (2009-12-30)

Known issues

• ipp008 - complete raid failure 2009.10.28 -- disks have been replaced and raid needs to be re-built. ipp008 has not been used for storage for some time, so we have not lost any vital data.
• ipp009 - several recent crashes with coincident complaints from CPU #3. note: we will swap CPU 1 and 3 and then stress-test this machine to see if the effect moves with the CPU
• ipp013 - dead fans, but it is not the fans themselves (tried replacing them). It appears to be the cabling that leads up to the fan modules that's gone bad.
• ipp014 - has a newer version of the Wave #1 motherboard which requires older kernel drivers or kernel 2.6.31.5.
• ipp016 - dead fans, but it is not the fans themselves (tried replacing them). It appears to be the cabling that leads up to the fan modules that's gone bad.
• ipp018 - suspect bad memory module (system is usable)
• ipp025 - suspect bad memory module (system is usable)
• ipp027 - suspect bad memory module (system is usable)
• ipp037 - fails to boot; serial device not detected by acs48, suspect motherboard
• ipp034 - Areca reported: 2009-11-17 13:42:23 Controller#1(PCI) H/W MONITOR DRAM 1-Bit ECC (online events logs states to Check the DRAM module and replace with new one if required.)

• ipp008, ipp046, ippdb02 - after mhpcc power outage (2009-11-27), the nodes fail to retain BIOS settings when power is totally lost. PDU outlets are powered off after shutdown procedure to prevent electrical surge when power resumes.

Old Issues

• random system crashes under heavy load of nodes, occasionally with a printk() of "do_IRQ: X.XXX" which appears to be caused by a hardware interrupt that does not have a handler (device driver) for it
  • these were only present on Wave #1 hardware
  • these seem to have largely stopped with the current forcedeth driver and kernel (early 2009?)

• ipp004 - disk bay #12 was dead, but started working again after the disk was reseated (ipp004 has device labels inverted, and it is suspected that previous re-seat attempts were addressing the wrong disk)

• cabinet 4 managed power strip was being overloaded. With the wave 3 delivery, we rebalanced the power load to avoid this problem.
• cab2pdu0 - cab2 down (2009-11-14) PDU sockets 5 & 9 reports on but PSU's on ipp005 & ipp009 not powered. Plan to return for repair.

• Entire IPP Production Cluster. MHPCC is currently experiencing a power outage. (2009-11-27T08:59)

Layout at MHPCC

  cabinet0: ipp020-21
  cabinet1: ipp012-013,037,008,014,016,018-019
  cabinet2: ipp004-007,015,009-011
  cabinet3: ipp023-029
  cabinet4: ipp030-036,017
  cabinet5: ipp038-043,ippdb00-02 (cab5con.ipp)
  cabinet6: ipp044-048,ippc00-09  (cab6con.ipp)
  cabinet7: ipp049-053,ippc10-19  (cab7con.ipp)