OpenVZ Forum


Home » General » Support » Dropped packet, source wrong veid
Dropped packet, source wrong veid [message #30580] Thu, 29 May 2008 08:27 Go to next message
ittec is currently offline  ittec
Messages: 41
Registered: February 2008
Member
Hi

Im having problems with a specific software: Plesk + Openads . I have 2 plesk runnings in separately VEs without any trouble. I have third container with Openads software installed, a openads frontweb, not a bd. Well, when the container start to receip some traffic the container stop to response. I don't know if is a problem with setup or something. I found this error in HN

Dropped packet, source wrong veid=1 src-IP=<ip container> dst-IP=83.50.166.85

when the VE fall. The other two Plesk are running perfectly but when i send some real traffic to third Ve with openads, it stops after some minutes.

Can I do some test or something to resolve it?

Thanks.


Re: Dropped packet, source wrong veid [message #30697 is a reply to message #30580] Wed, 04 June 2008 08:34 Go to previous messageGo to next message
ittec is currently offline  ittec
Messages: 41
Registered: February 2008
Member
After that I examined properly all messages of syslog I found souspicious things with relation with my problem.

The symptom is that some VE cannot answer all TCP paquets. First I had troubles with kmemsize, after some minutes, into Ve bash I received this message:

Cannot to allocate memory

or

Unable to fork

I read openvz wiki for kmemsize parameter and "lowmemory". By default, using vzsplite the kmemsize value assigned is not enough. I read lowmemory is the first 832 MB in i386 system. I assigned 100 MB to kmemsize for the problematic VE. After that, Ve not failed more. But, wasn't able to serve all traffic. Then I noticed in syslog the next

TCP: time wait bucket table overflow

I read that this bug was fixed in 2.6.18

http://bugzilla.openvz.org/show_bug.cgi

Now im thinking of increase /proc/sys/net/ipv4/tcp_max_tw_buckets_ub but...based on ... what? Default value is 16384. How affects this on hardware node? Are you agree to increase max_tw_buckets is properly solution?

Thanks
Re: Dropped packet, source wrong veid [message #30730 is a reply to message #30697] Thu, 05 June 2008 07:46 Go to previous message
ittec is currently offline  ittec
Messages: 41
Registered: February 2008
Member
After one day of changes, I can say that increase /proc/sys/net/ipv4/tcp_max_tw_buckets_ub was a good solution for my VE issue.

Thanks.
Previous Topic: VE start failed
Next Topic: 2.6.18-028stab053 doesn't find my NIC.
Goto Forum:
  


Current Time: Mon Nov 18 22:59:17 GMT 2024

Total time taken to generate the page: 0.02970 seconds