net.ipv6.route.max_size

net.ipv6.route.max_size

[Lukasz Trabinski]

Hello

IPv6 route table size is still growing. In stanard linux box configuration
with IPv6 BGP session (with ipv6 world table) and BGP application like Quagga,
we have problem like this:

2011/02/12 18:43:52 ZEBRA: netlink-cmd error: Cannot allocate memory,
type=RTM_NEWROUTE(24), seq=11124600, pid=0

This problem shows us that's all prefixes comes from BGP can't be install into 
FIB

IN BGP application we can see, that's we recive 4484 prefixes from ipv6 
neighbor:

cosmos> show bgp summary
Neighbor        V    AS MsgRcvd MsgSent   TblVer  InQ OutQ Up/Down State/PfxRcd
2001:1a68:0:c::1
                 4 24724  431945   70172        0    0    0 2d01h18m     4484


I suggest that net->ipv6.sysctl.ip6_rt_max_size should be increase.
Below _simple_ patch for this problem. I don't understand why in ipv4/route we 
have different method to count ip4_rt_max_size? My value 65536 in near future 
will be also wrong. I know also, that we can change it by sysctl, but I thing 
that it's not good way to fix it.


--- route.c~    2011-02-08 01:03:55.000000000 +0100
+++ route.c     2011-02-14 20:48:53.000000000 +0100
@@ -2708,7 +2708,7 @@
  #endif

         net->ipv6.sysctl.flush_delay = 0;
-       net->ipv6.sysctl.ip6_rt_max_size = 4096;
+       net->ipv6.sysctl.ip6_rt_max_size = 65536;
         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;

net.ipv6.route.max_size

[Lukasz Trabinski]

Hello

IPv6 route table size is still growing. In stanard linux box configuration
with IPv6 BGP session (with ipv6 world table) and BGP application like Quagga,
we have problem like this:

2011/02/12 18:43:52 ZEBRA: netlink-cmd error: Cannot allocate memory,
type=RTM_NEWROUTE(24), seq=11124600, pid=0

This problem shows us that's all prefixes comes from BGP can't be install 
into FIB

IN BGP application we can see, that's we recive 4484 prefixes from ipv6 
neighbor:

cosmos> show bgp summary
Neighbor        V    AS MsgRcvd MsgSent   TblVer  InQ OutQ Up/Down 
State/PfxRcd
2001:1a68:0:c::1
                 4 24724  431945   70172        0    0    0 2d01h18m     4484


I suggest that net->ipv6.sysctl.ip6_rt_max_size should be increase.
Below _simple_ patch for this problem. I don't understand why in 
ipv4/route we have different method to count ip4_rt_max_size? My 
value 65536 in near future will be also wrong. I know also, that we can 
change it by sysctl, but I thing that it's not good way to fix it.


--- route.c~    2011-02-14 20:26:12.000000000 +0100
+++ route.c     2011-02-14 20:30:04.000000000 +0100
@@ -2708,7 +2708,7 @@
  #endif

         net->ipv6.sysctl.flush_delay = 0;
-       net->ipv6.sysctl.ip6_rt_max_size = 4096;
+       net->ipv6.sysctl.ip6_rt_max_size = 65536;
         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;