Hello,
On Wed, 31 Oct 2001, Wayne wrote:
> According to the 2.2.19 kernel version of
> "net/ip_masq.h" (which the LVS patch ipvs-1.0.8
> does not change),
>
> /*
> * Linux ports don't normally get allocated above 32K.
> * I used an extra 4K port-space
> */
> #define PORT_MASQ_BEGIN 61000
> #define PORT_MASQ_END (PORT_MASQ_BEGIN+4096)
> The comment seems to imply that I could widen or
> move the range [PORT_MASQ_BEGIN, PORT_MASQ_END] as
> long as it starts above 32 k and ends below 64 k.
Yes. Even 2.4 sometimes selects 32768..61000
> Do you foresee any problems making this range bigger
> or perhaps moving it to another location above 32 k?
>
> Also the constant "PORT_MASQ_MUL" from
> "net/ipv4/ip_masq.c" appears to serve only as
> a check to make sure that the masquerading facility
> does not hog all the available memory, and that
> actually things would still work no matter how large
> it is or even if the checks using it are disabled
> altogether. Is this statement true?
We discussed this issue May-2000. It is in the HOWTO. By
multiplying this constant with the masq port range size you define
the connection limit for each protocol. Yes, this is
related to the used memory for masquerading and this is
a real limit but not for the LVS connections because they
are usually not limited by port collisions and LVS does not
check this limit.
> Thanks.
Regards
--
Julian Anastasov <ja@xxxxxx>
|