UrBackup Server 2.0.32 100% CPU usage

When my workstation is backing up to the server (Image Backup) the CPU on the server goes to 100%, creating a blocking problem with other services.

My server is an Proliant with a single Intel® Xeon® CPU L5420 @ 2.50GHz and 8GB ram running FreeNAS (UR Backup is running in a jail).

I have seen this same issue on two other systems I have setup in a similar way. All clients are on the local lan, not connected via the internet.

Any thoughts on how to de-prioritize the urbackupsrv process? anyone else seeing this issue?

On the server, under the settings -> advanced tab, do you have the box checked to be a background priority or not?

That is probably for workstations, not servers. My bad.

Yes, but that settings is for the clients? Full Text is “Run backups with background priority on the clients:

trying to use nice in a rc.d script, but its not working and server is now really slow. Getting 71.15 KBit/s was getting 100+ MBit/s on the activity page.

#!/bin/sh
. /etc/rc.subr
name=urbackupsrv
rcvar=urbackup_srv_enable
command=“/usr/local/sbin/${name}”
command_args=“run -d -g 104857600 -u root”
start_precmd=“cd /usr/local/var”
load_rc_config $name
run_rc_command “$1”
pidfile=“/var/run/${name}.pid”

start_cmd=“${name}_start”

urbackupsrv_start() {

    /usr/bin/nice -n 10 ${command} ${command_args}

}

I wonder if the jail is incurring some sort of performance penalty. Maybe try installing the server outside of a FreeBSD jail and see what the results are?

When I comment out the last few lines and eliminate the nice, I can you 100% of the server CPU to the point that other services start preforming poorly.

#!/bin/sh
. /etc/rc.subr
name=urbackupsrv
rcvar=urbackup_srv_enable
command="/usr/local/sbin/${name}"
command_args="run -d -g 104857600 -u root"
start_precmd="cd /usr/local/var"
load_rc_config $name
run_rc_command "$1"
pidfile="/var/run/${name}.pid"

#start_cmd="${name}_start"

#urbackupsrv_start() {

#       /usr/bin/nice -20 ${command} ${command_args}
#}

This is a backup running

But it looks like my issue with my first test was client side, after starting the sever with the rc.d script in my first post (with nice) I did not restart the client on my computer. Its now running at full speed, just need to test if its still blocking other process when hogging the CPU.

Great find!

no joy back to the drawing board, none of the above scripts where doing anything as they were not correct. once I got the correct order, and added an echo so I would know what was happening did i realize I had issues.

#!/bin/sh
. /etc/rc.subr
name=urbackupsrv
rcvar=urbackup_srv_enable

#start_cmd="${name}_start"
start_postcmd="${name}_poststart"
start_precmd="cd /usr/local/var"

command="/usr/local/sbin/${name}"
command_args="run -d -g 104857600 -u root"
pidfile="/var/run/${name}.pid"

urbackupsrv_start() {
        echo 'Starting "niced"'
        /usr/bin/nice --20 ${command} ${command_args}
}

urbackupsrv_poststart() {
        sleep 1
        echo 'Renicing'
        /usr/bin/renice -20 $(cat ${pidfile})
}

load_rc_config $name
run_rc_command "$1"

Problems

  1. the rc.d init system does not have permission to nice or renice :frowning:
  2. even when I renice urbackupsrv process to -20, it still blocks other process with its 100% CPU usage during image backups.

Bummer. Maybe try outside of the FreeBSD Jail?

Installed on FreeNAS so outside the jails really sounds like a bad idea, plus I’m start to consider this a bug. Should backing up 1 computer send the backup server to 100% CPU?

I have found a “work around” that pretty much sucks but is working. Under settings -> Server I have set “Total max backup speed for local network:”, this seems to limit the CPU leaving some idle cycles and my servers still working.

I can confirm that this issues is happening on two separate FreeNAS servers running 2.0.32 in a jail and that image backups send the CPU thru the roof.

Hehe, probably right. :wink:

I wonder if there is something else in the FreeNAS setup that is trying to also read/write disk changes at the same time that the image is being written (e.g. file system snapshots?).

Just to be clear, its NOT an IO issue its a CPU issue.

The server that I’m testing in my office is a Quad Core Xeon, with URBackup server running and idle the CPU is at 98% idle, when I initiate a image backup from a single 100% usage and its starves enough resources from the iscsi that it breaks my connections.

This is with a backup running from a single computer.

This is right after I stopped the backup.

I’m running urbackup 2.0.33 server in a jail on freenas 9.10.1.
I’m idling while backing up 30+ clients simultaneously. Granted they’re file backups.

I am NOT using ISCSI. So maybe that’s where the problem lies…

Can you try to run an image backup :), see what your CPU does? I have no issues with with file backup.

I use the FreeNAS as an ISCSI target, not involved in the URBackups setup. URBackups is running in a jail on the FreeNAS so it has direct access to my storage rather then a VM over NFS or ISCSI.

I only have IO issues with the VM’s using FreeNAS over ISCSI (which is most of my VM’S) when urbackupsrv is running a t 100% of the CPU. local IO on the freenas server is fine.

I’m running a full image backup of a W10 system right now.
The data is coming in at about 160Mbps, it’s on the LAN, and the CPU is averaging almost 97% idle. It also happens to be backing up 8 internet clients at the same time (file backup).

Are you using deduplication on the ZFS pool or encryption?
I only have compression running and rely on the urbackup deduplication.

The FreeNAS box here is running RAIDZ2 on 2 x 6 core AMD opterons with 48GB RAM and 15 x 3TB drives in groups of 5. Almost forgot, I have 2 x SSDs running as ZIL and L2ARC. Maybe that’s what helps…?

hippy1970 thanks! you pointed me in the right direction.

I setup both FreeNAS systems the same (vanila config). The one no-standard thing that I did do because it sounded good in my head was to use gzip level 9 compression on the URBackup datastore thinking that it would save me some space.

Well I switched back to LZ4 default and my CPU issues are gone.

1 Like

That’s great news.
Glad I could be of help.