|
Hello,
On Tue, 20 Nov 2001, Hendrik Thiel wrote:
> Hi All,
>
> I am using IP Virtual Server version 0.9.14 (size=4096)
>
> we have 6 Realservers each with these kind of active and inactive
> Connections
>
> -> RemoteAddress:Port Forward Weight ActiveConn InActConn
> -> server1:www Masq 1 68 12391
>
> today we reached a new peak (very fast, few minutes) from normaly 15Mbit/s
> to 30Mbit/s and the performance went down.
> Afterwards the following kernel messages (dmesg) showed up...
>
>
> IP_MASQ:ip_masq_new(proto=UDP): could not get free masq entry (free=31894).
> IP_MASQ:ip_masq_new(proto=UDP): could not get free masq entry (free=31894).
> IP_MASQ:ip_masq_new(proto=UDP): could not get free masq entry (free=31888).
It seems you are flooding some remote host from an internal
box, UDP protocol.
> I thought that the lvs and the masq table are two different things. What
> exactly
> does that mean? Is it possible that the fast and big amount of new
> connections
> did that?
www is TCP, you reach the UDP limit per destination (4096),
there are still free UDP ports for other destinations.
> another strange thing i figured out is the number of RX packets. I read
> something in the howto, but there it states that the TX packets were the
> problem. Unfortunatly we use eepro100 cards.
> Is it still a problem with lvs 0.9.14?
what kernel? Try to upgrade to latest. I remember for
such problem but forgot the driver.
> eth0:
> RX packets:2147483647 (2^31-1 with eepro)
> TX packets:1517722411
>
> any help appreciated. thanks in advance...
Regards
--
Julian Anastasov <ja@xxxxxx>
|
