Server unexpected reboot copying files

[netcsvg]

I'm new to FreeNAS and zfs. Our challenge:

We have about 500G worth of backup windows files, around 4G each on a local drive.
Goal is to copy those files to FreeNAS using Robocopy and iSCSI.

Every time we run the process, we get about 40 files copied (1+ hour into the process) before the FreeNAS server reboot by itself no errors logged.

Memory gets consumed well. MRU grows until it reaches the maximum and starts giving it back

Processor usage is almost nothing

Robocopy runs the 1G NIC at about constant 90%.


We have the newest firmware installed on the system, and hardware tests all look good.

We are currently running FreeNAS version 11, we were getting a similar issue with the prior version (9 something). It would crash at about 30 files copied

We have changed all hardware (except disks) and still getting the same issue.


Our config:

Dell R720xd
64GB ECC ram
NIC 2 port 1GB Intel
Disk Controller LSI SAS 9211 8i with JBOD firmware
Drives 12x 3G DeskstarNAS 7200rpms sata 3
Boot Drives 2x Intel 160G

FreeNAS 11.0-U2
single RaidZ3-0 vol
No tunables.
ISCI default values config.

any help would be appreciated...

Victor

 

[Chris Moore]

Why iSCSI?

 

[Stux]

Are there any ECC errors being reported in the IPMI event log?

What firmware is on the sas card?

With no errors being logged that implies a spontaneous restart. Either dodgy power, ram or CPU overheating.

Since the CPU is not very busy, could be a cooling failure? Or perhaps the heatsink is not applied. correctlt?

 

[netcsvg]

Why iSCSI?

it is the allowed protocol in our data network.

 

[netcsvg]

Are there any ECC errors being reported in the IPMI event log?

What firmware is on the sas card?

With no errors being logged that implies a spontaneous restart. Either dodgy power, ram or CPU overheating.

Since the CPU is not very busy, could be a cooling failure? Or perhaps the heatsink is not applied. correctlt?

the sas card has the 20 IT firmware

our UPS provides stable power, server is dual 1000W power suply, ram passed MemTest and self diagnostics. Also we have tried in another server (exact same config) with similar results.

It seems that we hit a limit somewhere, because it is always the same amount of information that gets transfer.

I'm thinking on using the SSDs as write cache and use a usb to boot...

 

[netcsvg]

it seems that the throughput of the raidz3 drive couldn't keep up with the network speed. Changing the config to two raidz2 vols fixed the issue. thank you all.

 

[Stux]

That would've doubled your IOPS

 

[Chris Moore]

I don't see why it would have rebooted instead of just running slower. If it works with two RAIDz2 vdevs but reboots with one RAIDz3 vdev, and there are no other changes, that just feels wrong. You might want to submit a bug report and get the developers to look into that.

 

[getum]

In such cases such as "copy long file names robocopy error", you may try a Long Path tool. I hope that helps.