- Joined
- Feb 6, 2014
- Messages
- 5,112
Assuming that it was MRU before and RR now, you should start seeing traffic on both NICs, unless there is something wrong with your iSCSI portal/target settings.
Assuming that it was MRU before and RR now, you should start seeing traffic on both NICs, unless there is something wrong with your iSCSI portal/target settings.
That means you probably have bad RAM, likely on memory bank 8 (whatever your hardware maps that to)I've got this messages in messages, what do they mean? Should I restart my freenas and/or the esxi host after the roundrobin multipath modify?:
View attachment 25656
That means you probably have bad RAM, likely on memory bank 8 (whatever your hardware maps that to)
You should check the iLO/out-of-band management logs on your server and hopefully it will identify the faulty DIMM for you by processor and slot
That means you probably have bad RAM, likely on memory bank 8 (whatever your hardware maps that to)
You should check the iLO/out-of-band management logs on your server and hopefully it will identify the faulty DIMM for you by processor and slot
You definitely have memory problems. I have seen the MCA errors in some of my servers. Some motherboards are more sensitive than others. I got some of those errors with mixed manufacturers of the installed DIMM's even though they all had the same spec. In any case, you can probably expect wonky things to keep happening until you make those MCA errors go away.
I'm sorry but what is MCA ??
Yeah, I should check the iLO because I got the error on memory again:
I'm also seeing that the memory isn't being released, it keeps getting consumed after the backups, but shouldn't it be released after the backup job (via iscsi) finishes?
In the error screenshot, it has "freenas MCA" as part of the text of the error message. I was abbreviating because I am a lousy typist. :)
Definitely a bad stick. Check the iLO logs and it should hopefully call out the stick.
No, the ARC will stay populated until it needs to replace it with more appropriate cached data. Free RAM is wasted RAM.
ipmitool sel elist
and comb through those lines to see if it identifies memory faults there.Tryipmitool sel elist
and comb through those lines to see if it identifies memory faults there.
For Intel NICs, anything dual-ported on the Pro/1000 series should work, but the older PT cards are less power-efficient. Look for the ET or newer series.
Helpful links:
https://forums.servethehome.com/ind...a-friends-don't-let-friends-buy-realtek.2663/
Comparison:
https://ark.intel.com/products/family/46827/Gigabit-Ethernet-Adapters#@Gigabit-Ethernet-Adapters
While it isn't any help in narrowing down the memory issue, that fact that the hardware watchdog is forcing hard resets on the system definitely points to "something is really wrong with the hardware"
Is there a "system event log" or similarly named in the BIOS? That might record the memory faults you need to trace down the bad DIMM.
Yes. ZFS will use (almost) all available RAM for the ARC, and will free it up if another process requests it.Is this normal? Just 340 MB of free RAM:
View attachment 25710
Yes. ZFS will use (almost) all available RAM for the ARC, and will free it up if another process requests it.
Scrubbing is basically an error-check on your pool and proactively looking for any data corruption. You definitely want these running routinely as well as SMART tests.ok, I have another question, what is the "scrub" for? I receive emails telling me that the scrub is running.
Thanks