Plesk and CPU usage

[shwonder]

Workaround was created only for cases when you have external DNS server. You can disable local DNS service on your server with Plesk and use external DNS.
Some days ago we created other workaround which fixes problem completely. Bind is configured to work on 1 CPU and it makes it do not waste CPU resources. You do not need to disable Bind on your server. In some days hotfix will be officially released. Please wait some days while it is completely will be tested by our QA department.

Best regards.

 

[Mertz]

EDIT: my apoligies if this seems off-topic.

This doesn't affect my company since we use external DNS servers and kill BIND under Plesk and we use Opterons, but..

I've always wondered why SW-Soft is using BIND on Win32. Why not use the built-in native and supported Microsoft DNS? I understand BIND gives a big benefit to have your management scripts easier to port over from the Linux product, but it seems to be the drawbacks are greater from using a buggy win32 port of dubious quality.

The MS DNS service is probably not as hefty as BIND running on native *nix but I can't think of a lot of things that get ported that are. Why not just use what's there and already regression tested by the maker of the OS?

 

[PaulC]

If I rememeber reading correctly, MS DNS support is coming in the next version.
Plesk 7 was mainly a fix to 6.5 in my eyes, and so not many new features were added or supported as the problems needed to be resolved.

 

[olivierfromars]

hi all
how to reload bind properly?
have I to open a shell window?
and type cd programfiles....?

 

[AbsolutelyFreeW]

I wouldn't like running MS dns, just my opinion, would rather have it *nix like even in future

 

[0wn4g3]

We are experiencing the same problem on our Windows Server 2003 box with Plesk 7. The stupid process called named.exe which refers to the Plesk Domain Service is taking up 90-100% of our CPU access. This is rediculous.

I can restart the Plesk Domain Service and it will fix it for a little while and then it acts up again. It has been so bad as of late that the box won't accept any RDC connections and it will load websites hosted on the box very slowly. Only a reboot of the machine temp fixes the problem. What the heck gives?

 

[truthmonkey42]

Possible Solution

I realize that this is years after this thread started, but we think we found a solution to this issue.

Apparently on multi-processor boxes BIND gets a little...'confused' if it's being told it only has one proc.

In the Service definition (under Services) the switch -n1 is the default. You can try (we just implemented this on a Plesk 7 box on Win2003 and so far so good (knock on wood))

Change the registry setting for the named service and change the switch to -nX

(X representing # of procs... 2 for a dual proc, 4 for a dualproc hyperthreaded machine)

Hope this helps someone out there!