"syncer" daemon taking more CPU time.

[abhilash_op]

Hi All,

In one of my HP server "syncer" daemon is taking excess amount of CPU time due to which the data transfer between servers has become slow, any remedy for this particular problem?.

Server OS : hp-ux 11.11

CPU TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
0 ? 547 root 152 20 2544K 508K run 15460:33 342.16 341.56 syncer


Is it safe to restart this daemon?.

Pls Reply fast

Abhi

[Ivan Krastev]

Check for changes in network cards - half duplex for example. This will cause syncer to consumer more cpu if you have NFS shares.


regards,
ivan

[Bill Hassell]

No. Do not kill this process as your system may become unstable. It is consuming a large amount of CPU time because of large I/O loads, particularly writes to the filesystems. Your buffer cache may be excessively large (multiple gigabytes) which will make the syncer run for longer periods. But syncer is not slowing down your system -- it is a symptom of massive I/O which was requested by your applications.

I would start by getting the patches up to date. Then analyze the current load using Glance to verify that the load is expected. If not, use the Glance data to revise application configurations and for databases, revise the query code and indexes.

[Steven E. Protter]

Shalom,

I think the first place i look when a critical system process is taking too much CPU is system use and patching.

What is your current patch level?

The June 2006 set is stable and may solve this and many other problems.

SEP

[abhilash_op]

Hi,

The patch level is realy old one

BUNDLE B.11.11 Patch Bundle
BUNDLE11i B.11.11.0306.1 Required Patch Bundle for HP-UX 11i, June 2003
FEATURE11-11 B.11.11.0209.5 Feature Enablement Patches for HP-UX 11i, Sept 2002
GOLDAPPS11i B.11.11.0306.4 Gold Applications Patches for HP-UX 11i, June 2003
GOLDBASE11i B.11.11.0306.4 Gold Base Patches for HP-UX 11i, June 2003
HWEnable11i B.11.11.0309.4 Hardware Enablement Patches for HP-UX 11i, September 2003


But is there any quick solution we can give for this issue, because the user is complaining that the data transfer between another server is slow. I have checked the second server, their CPU utlisation is normal.

Lancard speed is showing 100Mbs

# lanadmin -s 6
Speed = 100000000

[abhilash_op]

I could see my ignite network backup for this particular server is running. Will this be the Root cause for this high syncer CPU utilization?

[Bill Hassell]

> But is there any quick solution we can give for this issue, because the user is complaining that the data transfer between another server is slow.

It sounds like the single LAN card is running as fast as it can. A net_backup is a huge load too and if this server is where the data is stored, then, yes, syncer will be busy flushing data from the buffer cache. If the user is also sending data to this same server, that doubles the data flush rate.

However, LAN cards, especially slower cards like the 100 Mbit can't really keep modern disks busy at all. At best, a singloe 100 Mbit card can transfer perhaps 5 Mbytes/sec, about 1/4 the speed of normal disks.

Almost all performance issues require some study and the time to resolve issues very much depends on the tools. Classic Unix tools may take hours to figure out what is happening -- while Glance should pinpoint unusual processes or other resource usage in a few minutes.

[abhilash_op]

Hi,

I could resolve the problem with "syncer" daemon once i stopped the ignite backup and restarted the nfs.client. Now syncer is using very less CPU.

But right now the problem is with data transfer rate, there are 3 server mot01pa3,mot01da1 and mot01da2. User is complaining that the data transfer between mot01pa3 and mot01da1 takes longer time compared to data transfer between mot01pa3 and mot01da2. I am attaching my observations in the attachement.Pls suggest if you have any clue on this.

Thanks

Abhi