BigIron RX next-hop limitations
From FoundryWiki
Setting the nexthop cache setting of an RX line card too low may result in high CPU usage.
Contents |
[edit] Symptoms
When pinging IP addresses configured on an RX virtual interface start having abnormally high latency, the nexthop cache may be set too low. Although it may be the only symptom, it is without doubt advisable to have it set to a decent limit.
[edit] Checking the symptoms
As mentioned before, pinging an IP configured on the RX may show an abnormally high latency. The high latency only occurs when pinging from a device connected to the line card itsself.
Additionally, you can check the nexthop cache setting of a linecard, like this:
SSH@rx8#rconsole 1 Remote connection to LP slot 1 established Press CTRL-X or type 'exit' to disconnect it LP-1>show ip nexthop Paths Total Free In-use 1 2816 2158 658 2 512 512 0 4 512 512 0 8 256 0 256 (lots of other nexthop information)
As you can see, this linecard has no additional free entries left in the eightpath section. According to a Foundry TAC engineer, this may the cause of the high latency and some other unexplainable problems.
[edit] Configuring a higher limit
Configuring a higher limit involves changing the cam-partition, which obviously requires a reload. To raise the limit from the default 256 to for example 1024, enter the following commands:
SSH@rx8#conf t cam-partition next-hop 2816 1024 1024 1024
This will change the default limits for 2, 4 and 8 paths to 1024. The 2816 value is the default value for 1 path.
As you leave configuration mode, the box will notify you about the reload required.
