amd64 (Xeon 3450) with 8GB of RAM and four 250GB SATA drives in a RAID 1+0. Upgraded it to 9.2-RELEASE the other day, hoping to fix some of (what I think) are ZFS related problems.
I found zfs-stats in ports.
http://pastebin.com/CBYLGp56
The machine is also being backed up with BackupPC.
I have read that rsync and ZFS do not play nice together, and Maildir and ZFS do not play nice together.
This is a mailserver for a company that has over 1k users on it. (qmail and dovecot.)
I am looking for a way to make it more stable. What happens is the entire machine is unresponsive and only a "walk over and power down" will get it to work again. Apparently this has been happening to them for a few months now, but no one ever said anything and the exiting admin neglected to bring that up either.
Thinking about moving the whole thing to a similar machine with UFS if I can not get this stable.
I see there are a ton of "tuning the ARC cache" but I can not find out where/when you decide what/why to set the values to.. most people seem to randomly choose values and "hope for the best."
Looking to not build a future on hope.
Suggestions for what I could to do determine if ZFS (ARC or something else) is my problem. And then to figure out what to do with the problem.. Again my understanding is that "ZFS is the problem" in this situation.
I found zfs-stats in ports.
Code:
zfs-mon:
ZFS real-time cache activity monitor
Seconds elapsed: 161
Cache hits and misses:
1s 10s 60s tot
ARC hits: 167 255 287 330
ARC misses: 0 0 1 6
ARC demand data hits: 42 53 47 59
ARC demand data misses: 0 0 0 0
ARC demand metadata hits: 125 202 238 268
ARC demand metadata misses: 0 0 0 4
ARC prefetch data hits: 0 0 0 0
ARC prefetch data misses: 0 0 0 0
ARC prefetch metadata hits: 0 0 2 3
ARC prefetch metadata misses: 0 0 0 2
ZFETCH hits: 800 1248 1159 1295
ZFETCH misses: 34 49 85 224
Cache efficiency percentage:
10s 60s tot
ARC: 100.00 99.65 98.21
ARC demand data: 100.00 100.00 100.00
ARC demand metadata: 100.00 100.00 98.53
ARC prefetch data: 0.00 0.00 0.00
ARC prefetch metadata: 0.00 100.00 60.00
ZFETCH: 96.22 93.17 85.25 zfs-stat -a:http://pastebin.com/CBYLGp56
egrep -v \# /boot/loader.conf:
Code:
zfs_load="YES"
ahci_load="YES"
geom_mirror_load="YES"
accf_http_load="YES"
accf_data_load="YES"The machine is also being backed up with BackupPC.
I have read that rsync and ZFS do not play nice together, and Maildir and ZFS do not play nice together.
This is a mailserver for a company that has over 1k users on it. (qmail and dovecot.)
I am looking for a way to make it more stable. What happens is the entire machine is unresponsive and only a "walk over and power down" will get it to work again. Apparently this has been happening to them for a few months now, but no one ever said anything and the exiting admin neglected to bring that up either.
Thinking about moving the whole thing to a similar machine with UFS if I can not get this stable.
I see there are a ton of "tuning the ARC cache" but I can not find out where/when you decide what/why to set the values to.. most people seem to randomly choose values and "hope for the best."
Looking to not build a future on hope.
Suggestions for what I could to do determine if ZFS (ARC or something else) is my problem. And then to figure out what to do with the problem.. Again my understanding is that "ZFS is the problem" in this situation.
