[PVE-User] High CPU and Memory utilzation

[Gilberto Nunes]

Hello PVE guys...

I have one host, which is a Dell PowerEdge R430, with 12 cores and 48 GB of
memory.
The disk are all SAS 10K RPM.
I have just one KVM VM with Ubuntu 14.04, which is our Zimbra Mail Server.
The qcow2 image is reside in another server, also a Dell PowerEdge R430,
with 15 GB of memory. This server, act like a Storage, a GlusterFS Server,
set up like NFS.
There's 3 SAS disk 10K RPM. Between this to servers, I have three 1 GB NIC
set as bond0, with Bonding Mode: load balancing.
This the iperf3 output:

iperf3 -c storage100
Connecting to host storage100, port 5201
[  4] local 10.1.1.140 port 36098 connected to 10.1.1.180 port 5201
[ ID] Interval           Transfer     Bandwidth       Retr  Cwnd
[  4]   0.00-1.00   sec   309 MBytes  2.60 Gbits/sec  101    386 KBytes

[  4]   1.00-2.00   sec   250 MBytes  2.10 Gbits/sec   20    387 KBytes

[  4]   2.00-3.00   sec   247 MBytes  2.07 Gbits/sec   20    390 KBytes

[  4]   3.00-4.00   sec   306 MBytes  2.56 Gbits/sec    0    390 KBytes

[  4]   4.00-5.00   sec   314 MBytes  2.63 Gbits/sec   13    390 KBytes

[  4]   5.00-6.00   sec   314 MBytes  2.63 Gbits/sec    0    390 KBytes

[  4]   6.00-7.00   sec   295 MBytes  2.48 Gbits/sec    0    390 KBytes

[  4]   7.00-8.00   sec   280 MBytes  2.35 Gbits/sec   90    468 KBytes

[  4]   8.00-9.00   sec   297 MBytes  2.49 Gbits/sec    0    468 KBytes

[  4]   9.00-10.00  sec   304 MBytes  2.55 Gbits/sec  142    485 KBytes

- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bandwidth       Retr
[  4]   0.00-10.00  sec  2.85 GBytes  2.45 Gbits/sec  386             sender
[  4]   0.00-10.00  sec  2.85 GBytes  2.44 Gbits/sec
 receiver

iperf Done.

Sometimes, and this is more often than I wish, I run into high cpu and
memory utilization. The CPU reach %CPU 250-300 and %MEM reach 77-80.
​
And I don't exactly this is the cause of slow access to disk, but in fact,
I get more than 50 MB/s in diskwrite. That is showing to me, from Disk IO
rrdtool graphics in PVE web console.
In this period of time, the Zimbra Mail Server has slow access and my users
complain a lot of time...

I had make some research in Google and try improve performance in glusterFS.
Here is something I working on:

gluster vol info

Volume Name: storage
Type: Distribute
Volume ID: 183b40bc-9e1d-4f2c-a772-ec8e15367485
Status: Started
Number of Bricks: 1
Transport-type: tcp
Bricks:
Brick1: storage100:/data/volume
Options Reconfigured:
performance.write-behind-window-size: 1024MB
performance.cache-size: 1GB
performance.io-thread-count: 16
performance.flush-behind: on
performance.cache-refresh-timeout: 10
performance.quick-read: off
performance.read-ahead: enable
network.ping-timeout: 2
performance.cache-max-file-size: 4MB
performance.md-cache-timeout: 1
nfs.addr-namelookup: off
performance.client-io-threads: on
performance.nfs.io-cache: off
performance.cache-priority: *.qcow2:1,*.raw:1


Is there some other thing I can do improve performance?

I'll appreciated any advice!...


Thanks a lot.