8.2.0 BETA3 - kernel panic

[StephenFry]

I'm reverting back to 8.0.4 or 8.2.0 BETA2 for the moment, this BETA3 is not working well:

Dump failed. Partition too small.
Automatic reboot in 15 seconds - press a key on the console to abort
panic: kmem_alloc(32768): kmem_map too small: 1603923968 total allocated
cpuid = 0
Uptime: 1h45m19s

It then fails to actually reboots and stays frozen.

 

[ProtoSD]

Are you running x86 or x64?

 

[StephenFry]

x64, all the specs for my test rig are in my sig and are current.

edit: this is a reproducible error whenever I copy a fair amount of data. During the copy it will crash at anywhere from 75-200GB.

The other builds have been stable so far.

edit2: because I'm stubborn, I upgraded beta2 once again to beta3 and have so far found three issues:
-this kernel panic
-the 'advanced mode' button in CIFS doesn't work
-in 'view disks' my drives don't show their serial nr.

 

[William Grzybowski]

About the panic... do you have anything under System -> Tunables?

I think you do, delete anything related to zfs or kmem... kmem_size, arc_max, etc

 

[StephenFry]

About the panic... do you have anything under System -> Tunables?

I think you do, delete anything related to zfs or kmem... kmem_size, arc_max, etc

I do, indeed. I haven't added it in myself, though.

Deleted vm.kmem_size 1536M
Deleted vfs.zfs.arc_max 1024M
I kept kern.ipc.nmbclusters 50000 (or should that go as well?)

Will start a new copy of data

 

[William Grzybowski]

You need to reboot so it can take effect

 

[StephenFry]

You need to reboot so it can take effect

Already did, and copying data. Will report back!

 

[StephenFry]

Well, that was quick :(

panic: kmem_malloc(131072): kmem_map too small: 1608253440 total allocated
cpuid = 1

and then it freezes and needs a hard reboot :(

edit: rebooted and now it has no configured LAN interface...
edit2: ignore that last message, another reboot solved that one.
(still, it seems beta3 really doesn't like my machine!)

 

[gcinini]

Same problem here.

FreeNAS-8.2.0-BETA3-x64
HP ProLiand Mini Server
2x 1GB DDR3
1x WD 250GB
3x WD Green 1.5TB RAIDZ1

Did not have problems copying medium and small files, but when copying large files it Panics with the kmem_malloc(131072) error.

The file I am using to test is a 13GB file.

I see that I have the vm.kmem_size entry on Tunables, Generated by autotune. Will turn autotune off and remove those entries. Will report back.

 

[gcinini]

Turned off Autotune and deleted two entries on Tunables:
Deleted vm.kmem_size
Deleted vfs.zfs.arc_max

I kept kern.ipc.nmbclusters entry.

Just rebooted, confirmed the entries where not there anymore.

And I could copy the 13GB file with no problems. Performance was not great, as it would pause at every few seconds.

Also, the problem would not happen with all large files, but with some specific ones.

Final note: I am using Windows 8 CP to copy the files from an external USB drive to the FreeNAS box over GBit Ethernet.

 

[gcinini]

OK, so I had the problem again. Downgraded to 8.0.4 and had the same problem.

I now turned compression off on all volumes and am trying again. I am thinking it might have to do with lack of memory.

 

[TravisT]

I'm having this same problem, and was wondering if you had figured anything out. I was running BETA3 when I first noticed the problem. I upgraded to RC1 hoping it would solve the problem. Unfortunately, it didn't.

I'm running this on ESXi 5.0 as a VM. The VM had 4GB RAM allocated. I upped it to 5GB yesterday and am still having the problem. I upped it to 8GB just now to see if that would make a difference. Any suggestions would be appreciated!

 

[cyberjock]

Personally, I found my FreeNAS machine worked best with 6GB of RAM with only 1TB of hard drive space allocated. Of course, I experiment with it a lot and I do alot of things you aren't supposed to do just to see what happens and learn from it.

 

[TravisT]

Thanks, but I'm planning on having my freenas serve up to 15 disks, with the smallest physical disk size being 1TB. I need something capable of reliably working with large storage pools.

Is anyone else having these problems? Something else I noticed is in my daily run output, I'm getting a 3ware RAID controller alarm occasionally. I don't have a 3ware RAID controller in my server, nor am I running RAID (other than raidz inside of freenas). Not sure if these could be related in any way.