Forum Discussion

RBS_79242's avatar
RBS_79242
Icon for Nimbostratus rankNimbostratus
Jun 23, 2013

Persistence Fault

Hi all,

 

I have an issue in that the F5 is refusing to add a persistence record. I have logged a fault with F5 support but it's not getting much traction so I thought I would ask in here.

 

The lines in my iRule look like this:

 

persist add uie $serverCRC

 

if {$debugDebug}{log local0. "MCOOK_RESD: [LB::server name] - [IP::client_addr]:[TCP::client_port] Lookup $serverCRC Node: [IP::server_addr] as: [persist lookup uie $serverCRC node]."}

 

 

and in the log I get the corresponding entry:

 

 

Jun 23 11:57:26 local/tmm info tmm[5202]: Rule med_cookie_persist_v2 : MCOOK_RESD: pool-prd21-www.govt.nz_http 10.204.204.125%540 80 - 122.57.XXX.109:57584 Lookup 015b95ec25b9c18260cdd5290185afa2 Node: 10.204.204.125%540 as: .

 

 

As you can see the LOOKUP has failed, resulting in no output. Because the "persist add" failed to execute.

 

 

This command is working fine for other Cookies, here is a look at the persistence table:

 

 

PERSISTENT CONNECTIONS

 

| Mode universal Value b264cc9f5482e9bec0260a205b14ff1f

 

| virtual XXX.XXX.228.178:http node 10.204.204.123%540:http age 9sec

 

| Mode universal Value acb7e15ed74cafcfc7e1a154194cfe75

 

| virtual XXX.XXX.228.178:http node 10.204.204.124%540:http age 16sec

 

| Mode universal Value 80d183bc27faf1a99dba5d05515f91e8

 

| virtual XXX.XXX.228.178:http node 10.204.204.126%540:http age 43sec

 

 

Now this is an intermittent fault. What I have discovered is, if I delete the persistence table with b persist virtual [name] delete the fault could completely go away. It could also move to another node refusing to persist. At the moment, (and for the last two days), the F5 is refusing to create a persistence record for that value. I am confident that if I delete the persistence table for the virtual that it will either (a) move to another node, or (b) completely disappear. The issue with it completely disappearing is that eventually it starts happening again as end nodes are taken off line.

 

 

Is there a way to get any more debugging information out of the persist command? I would really like to know why it is failing to create an entry.

 

 

Thanks.

 

 

Rowan

 

 

 

1 Reply

  • i am not aware a known issue. anyway, just wondering if problem still persists after disabling cmp.

     

     

    sol7751: Overview of Clustered Multi-Processing (9.x - 10.x)

     

    http://support.f5.com/kb/en-us/solutions/public/7000/700/sol7751.html