MrD1234

So much memory in a 32-bit environment?

attic? i hope it's cooled or you live in the north somewhere... my attic is 100+ degrees in the summer

yes thats true ipmi / kvm is a very nice feature

That 133 Mhz 64bit card is only running at 33Mhz 32 bit (i.e. an eight of its rated performance)

Double confirm, something causes the "upgradepkg" script to die (i.e <defunct>) You can telnet into the box and kill the process and the rest of the system comes up. All the vmtools is installed, so I am not quite sure what the issue is. Perhaps there are some remnants from a previous vmtools package?

Thanks! There is something weird going on for sure. I added another disk to the LSI 9211 controller, added it as a raw virtual disk and mapped it into Unraid. It completed the re-construction successfully, but when I checked the array after I got home, I saw the same errors on a different physical disk. I am wondering if there is some kind of spin up delay that the controller / driver is not waiting long enough and the driver thinks it is a physical error.

ok that makes sense. By "clear" i mean simulate replacing the drive I don't believe the drive is bad. I think it's a ESXi issue.

I had 2 read / 2 write errors occur on a disk. The statistics for that drive do not update but writes to that drive are still allowed to proceed via a samba share on a user share. Jul 6 23:26:13 Tower2 kernel: sd 1:0:2:0: [sdc] Device not ready Jul 6 23:26:13 Tower2 kernel: sd 1:0:2:0: [sdc] Result: hostbyte=0x00 driverbyte=0x08 Jul 6 23:26:13 Tower2 kernel: sd 1:0:2:0: [sdc] Sense Key : 0x2 [current] Jul 6 23:26:13 Tower2 kernel: sd 1:0:2:0: [sdc] ASC=0x4 ASCQ=0x2 Jul 6 23:26:13 Tower2 kernel: sd 1:0:2:0: [sdc] CDB: cdb[0]=0x28: 28 00 00 04 00 40 00 00 08 00 Jul 6 23:26:13 Tower2 kernel: end_request: I/O error, dev sdc, sector 262208 Jul 6 23:26:13 Tower2 kernel: md: disk2 read error Jul 6 23:26:13 Tower2 kernel: handle_stripe read error: 262144/2, count: 1 Jul 6 23:26:14 Tower2 kernel: sd 1:0:2:0: [sdc] Device not ready Jul 6 23:26:14 Tower2 kernel: sd 1:0:2:0: [sdc] Result: hostbyte=0x00 driverbyte=0x08 Jul 6 23:26:14 Tower2 kernel: sd 1:0:2:0: [sdc] Sense Key : 0x2 [current] Jul 6 23:26:14 Tower2 kernel: sd 1:0:2:0: [sdc] ASC=0x4 ASCQ=0x2 Jul 6 23:26:14 Tower2 kernel: sd 1:0:2:0: [sdc] CDB: cdb[0]=0x2a: 2a 00 00 04 00 40 00 00 08 00 Jul 6 23:26:14 Tower2 kernel: end_request: I/O error, dev sdc, sector 262208 Jul 6 23:26:14 Tower2 kernel: md: disk2 write error Jul 6 23:26:14 Tower2 kernel: handle_stripe write error: 262144/2, count: 1 Jul 6 23:26:14 Tower2 kernel: md: recovery thread woken up ... Jul 6 23:26:14 Tower2 kernel: md: recovery thread has nothing to resync Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] Device not ready Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] Result: hostbyte=0x00 driverbyte=0x08 Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] Sense Key : 0x2 [current] Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] ASC=0x4 ASCQ=0x2 Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] CDB: cdb[0]=0x28: 28 00 00 00 00 c0 00 00 08 00 Jul 6 23:26:15 Tower2 kernel: end_request: I/O error, dev sdc, sector 192 Jul 6 23:26:15 Tower2 kernel: md: disk2 read error Jul 6 23:26:15 Tower2 kernel: handle_stripe read error: 128/2, count: 1 Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] Device not ready Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] Result: hostbyte=0x00 driverbyte=0x08 Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] Sense Key : 0x2 [current] Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] ASC=0x4 ASCQ=0x2 Jul 6 23:26:15 Tower2 kernel: sd 1:0:2:0: [sdc] CDB: cdb[0]=0x2a: 2a 00 00 00 00 c0 00 00 08 00 Jul 6 23:26:15 Tower2 kernel: end_request: I/O error, dev sdc, sector 192 Jul 6 23:26:15 Tower2 kernel: md: disk2 write error Jul 6 23:26:15 Tower2 kernel: handle_stripe write error: 128/2, count: 1 Jul 6 23:37:01 Tower2 crond[1148]: ignoring /var/spool/cron/crontabs/root- (non-existent user) I've read the system should be in a read only mode. This feels like a bug. Also given this is a test array, what is the best way to "clear" the error without replacing the drive? EDIT -- and the parity check buttons are gone

Curious how much you paid for those. I have dozens of sata cables in a box, that I've collected over time. Every motherboard I get comes with 6 or 8 cables and I never use them.

Theoretically gigabit tops out at 125MB/s Gigabit on a good switch tops out at 110MB/s. not sure what you can use the last 7-23 MB/s for....

a gigabit controller will saturate the PCI bus.

It supports 256 drives via SAS expanders, theoretically with 10 unraid VMs that's a lot of future proofing

Like mine

aren't media (i.e. dvd / photos / music) storage drives idle most of the time... unless people spend more than 8-10 hours in front of the idiot box.

The expander only uses the pcie slot for power. If you're reasonably competent with a soldering iron, you could probably order a pcie connector like this one and solder a molex pigtail to the proper pins. You also need to assert the power good signal for the card to run.

Correct. However, the board does need a "power good" signal. Some boards have a tendency to not assert that signal if the bios scan does not detect a card. The newer Chenbro card based on the LSI chipset might be a better alternative since it gets power from the molex connecter and only uses any slot for structural support http://www.chenbro.com/corporatesite/products_detail.php?sku=187

Thanks! Yes one definitely does not spend $1k on a full feature RAID card for UnRAID BTW, I believe both the chenbro and Intel cards use a molex connector to power the board and can use any slot. Additionally there are "mezzanine" form factors of those cards too and you may be able to custom mount them.

You actually need both. The 4 port is an SAS HBA the second card is an SAS expander. The HBA card is what the OS talks to. The SAS expander is like a network switch for SAS / SATA drives. Actually the HP card is 36 ports, but in practice you can not use the external port and you have to sacrifice one or two ports to link to the HBA card. The LSI HBA is nice if you get the 8i (2 sas port) model you can dual link to the expander for double the bandwidth. There are also Intel and Chenbro models for the SAS expander but the Intel model only supports 20 drivers after linking and the Chenbro is mucho expensivo.

Yes. Try this instead http://www.newegg.com/Product/Product.aspx?Item=N82E16816118114&cm_re=lsi_9211-_-16-118-114-_-Product and http://www.newegg.com/Product/Product.aspx?Item=N82E16816401184&nm_mc=OTC-Froogle&cm_mmc=OTC-Froogle-_-Hard+Drive+Controllers+/+RAID+Cards-_-Hewlett-Packard-_-16401184 That case is only going to accommodate 3 of those bays as stated on the website. I would seriously not constrain the design with that case.

The biggest factor I have seen is switch quality. I have never seen > 60 MB on consumer hardware on a gigabit network. I replaced my switches with a hp switch and always see 90-100MB speeds to a pc and a linux server. My unraid server is VMed so I see speeds much less (12-25MB) write. I am waiting for /dev/vdx support.

No emhttp segfault here -- I had already set spin down to never. My SAS card controls the spin down. I did get beta5 to hard lock by writing large files from Windows 7 64 over gigabit (22MB/s) and simultaneously hit refresh on the /Main page. I was not able to capture anything and had to kill the VM.

use this http://www.newegg.com/Product/Product.aspx?Item=N82E16813182235

The Post / redirect / get pattern is the pattern that solves this problem. http://en.wikipedia.org/wiki/Post/Redirect/Get

The errors are USB related, but I thought the issue was a network issue. I would look to see if the USB driver and the network driver are sharing the same interrupt.

Ok good I am seeing this in the beta too with 500MB drives. The order changes. I was thinking this was related to being inside a VM (qemu-kvm).