High CPU load on modules

[waketech]

Hey there,

Has anybody experienced high CPU load on 24p modules in a 5400?  The specific modules we're noticing it on is J8706A and J8702A.

A bit more info:

We have a 5412 as our core router.  It's populated with 2 24p SFP modules and 5 24p Gig-T modules.  If we run 'sh cpu' everything looks fine.

3 percent busy, from 300 sec ago
1 sec ave: 1 percent busy
5 sec ave: 1 percent busy
1 min ave: 1 percent busy


Task usage for last 2 sec
 % CPU | Description
-------+--------------------------
  98.0 | Idle
   0.7 | Sessions & I/O
   0.7 | SNMP
   0.7 | IP Host/Routing

However, if we do 'sh cpu slot a' it shows pretty high CPU load

slot a:
-------
67 percent busy, from 300 sec ago
1 sec ave: 90 percent busy
5 sec ave: 88 percent busy
1 min ave: 89 percent busy

That was run at 6:30am when we have very little traffic on the network. Here's the port utilizations:

Status and Counters - Port Utilization

                                 Rx                           Tx           
 Port      Mode     | --------------------------- | ---------------------------
                    | Kbits/sec   Pkts/sec  Util  | Kbits/sec  Pkts/sec   Util
 -------- --------- + ---------- ---------- ----- + ---------- ---------- -----
 A1       .         | 0          0          0     | 0          0          0    
 A2       1000FDx   | 5664       148        00.56 | 5616       226        00.56
 A3       1000FDx   | 5104       41         00.51 | 5176       84         00.51
 A4       .         | 0          0          0     | 0          0          0    
 A5       1000FDx   | 26872      3524       02.68 | 32640      3410       03.26
 A6       .         | 0          0          0     | 0          0          0    
 A7       1000FDx   | 0          0          0     | 0          0          0    
 A8       1000FDx   | 0          0          0     | 0          0          0    
 A9       1000FDx   | 5000       33         00.50 | 5120       77         00.51
 A10      1000FDx   | 5232       51         00.52 | 5264       136        00.52
 A11      1000FDx   | 4968       0          00.49 | 5048       47         00.50
 A12      1000FDx   | 5192       44         00.51 | 5216       134        00.52
 A13      1000FDx   | 5304       73         00.53 | 5264       159        00.52
 A14      1000FDx   | 5160       33         00.51 | 5240       120        00.52
 A15      1000FDx   | 5152       24         00.51 | 5192       152        00.51
 A16      1000FDx   | 5520       181        00.55 | 6512       294        00.65
 A17      1000FDx   | 5328       154        00.53 | 5224       134        00.52
 A18      1000FDx   | 5184       43         00.51 | 5176       130        00.51
 A19      1000FDx   | 5168       57         00.51 | 5264       139        00.52
 A20      1000FDx   | 5160       52         00.51 | 5584       155        00.55
 A21      1000FDx   | 5256       95         00.52 | 5312       176        00.53
 A22      1000FDx   | 5256       64         00.52 | 5792       196        00.57
 A23      1000FDx   | 5424       97         00.54 | 5208       181        00.52
 A24      .         | 0          0          0     | 0          0          0    

We'll witness similar numbers throughout the day.  The other modules aren't as high (56, 51, 37, 10, 36; 5min avg) but they will climb a bit throughout the day...  We usually see 3 or 4 modules consistently over 70% usage (at 5min avg) with 2 SFP modules >90%.

I've tried running 'showtaskusage' from inside testmode, but it didn't show anything useful...  Here's a snippet of one:

slot e:
-------
                |  Recent |  %  |  Total  | Time Since |   Times  |   Max
 Task Name      |   Time  | CPU |  Time   |  Last Ran  |    Ran   |  Time
----------------+---------+-----+---------+------------|----------|-------
     mIpAdMUpCt |   53 ms |   2 |   114 ms|      11 ms |       44 |  12 ms
       eDevIdle |  460 ms |  25 |   901 ms|       2 ms |     4843 | 795 us
       eDrvPoll |   79 ms |   4 |   465 ms|      44 ms |       50 |  50 ms
     tDevPollRx |  875 ms |  47 |     2 s |       1 ms |     2460 |  10 ms
     tDevPollTx |  185 ms |  10 |   360 ms|       3 ms |     2417 | 800 us
     mPmSlvCtrl |   53 ms |   2 |   113 ms|     320 ms |        9 |  15 ms
     mPoeMgrCtl |   65 ms |   3 |   132 ms|     740 ms |      379 |   1 ms

Coming back to the original question, has anybody else noticed excessively high CPU usage on individual modules?