This post is more than 5 years old

5 Posts

18653

November 17th, 2012 06:00

pe1900 perc 5i can't enable write back

I replaced a failed 5iR with a reconditioned perc 5i from ebay. The system is a pe1900 (I) with  a single 70Gb dell SAS.

On initial boot there were errors about the battery charge - I left the system to charge.

On booting again I did ctrl-R  and initialized the disk as RAID0 with default settings + "write back" (this was not shown as default) .

No errors or warnings this time.

I did (is this right ?) to get out of the virtual disk manager and to reboot

Then I installed Fedora 17 from a CD with out any obvious problems.

However on closer inspection

the system boot log concerning the drive says " Assuming drive cache: write through"

here's the rest :

 megasas: 00.00.06.14-rc1 Fri. Jan. 6 17:00:00 PDT 2012
 megasas: 0x1028:0x0015:0x1028:0x1f03: bus 2:slot 14:func 0
 megaraid_sas 0000:02:0e.0: PCI IRQ 142 -> rerouted to legacy IRQ 18
 megasas: FW now in Ready state
 megasas_init_mfi: fw_support_ieee=0
 megasas: INIT adapter done
 scsi2 : LSI SAS based MegaRAID driver
 megaraid_sas: fw state:c0000000
 megasas: fwstate:c0000000, dis_OCR=0
 scsi 2:0:4:0: Direct-Access     MAXTOR   ATLAS10K5_073SAS BP00 PQ: 0 ANSI: 5
 scsi 2:2:0:0: Direct-Access     DELL     PERC 5/i         1.00 PQ: 0 ANSI: 5
 142082048 512-byte logical blocks: (72.7 GB/67.7 GiB)
 Test WP failed, assume Write Enabled
 Asking for cache data failed
 Assuming drive cache: write through
 sd 2:2:0:0: Attached scsi generic sg1 type 0
 Test WP failed, assume Write Enabled
 Asking for cache data failed
 Assuming drive cache: write through
  sda: sda1 sda2
 Test WP failed, assume Write Enabled
 Asking for cache data failed
 Assuming drive cache: write through
 Attached SCSI disk

For want of a better idea, I repeated this process and got the same result.

Oh and hdparm gives the following:

/dev/sda:
SG_IO: bad/missing sense data, sb[]:  70 00 05 00 00 00 00 0b 00 00 00 00 20 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 multcount     =  0 (off)
 readonly      =  0 (off)
 readahead     = 256 (on)
 geometry      = 8844/255/63, sectors = 142082048, start = 0

I don't really know what this means and google gives nothing I can understand about the SG_IO line.

But I did leave the Read Policy at default "no read ahead" in the VDM.

I guess next I will try the "restore factory defaults" option in the VDM.

I am a software guy with a woeful hardware knowledge, can does any of this raise any flags for anyone ? Am I missing some step (quite possible) ? Are there any other diagnostics I can try ?

And just to reiterate : I did initially get an amber light and a battery not charged warning - but have not seen anything since.

5 Posts

November 17th, 2012 15:00

After doing some measurements I can confirm that "write back" is actually working after all.

Seems like this is opaque to the OS - which makes sense I suppose.

Slightly curious why we get "Write Cache is Disabled" in the 32bit diagnostics though.

These figures give quite a nice graph for those who are into that kind of thing (note the spike at data sent = 2 x bufsize)

Data Sent(Mb) vs Transfer Rate(Mb/s)

601 347.398843930636

611 391.666666666667

621 363.157894736842

641 335.602094240838

661 277.731092436975

681 302.666666666667

701 273.828125

721 256.583629893238

741 242.156862745098

761 232.012195121951

781 224.425287356322

801 216.486486486486

901 191.295116772824

1001 175.922671353251

1101 173.11320754717

1201 170.354609929078

1401 171.901840490798

2001 196.176470588235

11 Legend

 • 

16.3K Posts

November 17th, 2012 07:00

The sense code says illegal operation/unsupported command ... at first glance the Atlas SAS drive should be ok, although the firmware is out of date (latest is BP05) ... however, I also doubt that is the cause, especially on an [older] PERC 5.

The system will not enable the cache until the battery reaches a certain level ... and the battery will only charge when the disk is in use.  If it has been running for more than several hours, and it still does this, then there may be another issue with the PERC - with its memory or its battery.

You might run 32-bit diagnostics on the controller/disk.  One of the Dell guys better versed in Linux could jump in here with additional ideas working with your OS - running OMSA, controller settings, etc.

5 Posts

November 17th, 2012 10:00

Thanks for that tip - I am running the 32B diagnostics extended test now....

5 Posts

November 17th, 2012 11:00

Ok so the diagnostics give no errors (apart from a couple of irrelevant cached ones)

Also under

Custom Tests->hard drive->configuration

I see

Write Cache....: Write Cache is Disabled

which I guess just confirms just that.

No Events found!

Top