Controllerless Networks

Reply
Occasional Contributor II

MDNS Memory over subscribed

Hi,

I've been going through our controllers (7240's and 7220's) (AOS 6.5.1.9) and have noticed that our mdns memory is over subscribed. I've seen it as high as a gig of memory over subscribed. Has anybody else seen this issue. I've had some complaints from users telling me their video streaming has cut out on them. One inparticular is a Solstice Pod. Here is a screen shot of memory over subscribed. 

 

#show memory debug | include over

arci-cli-helper : 34/100M (cur/typical) 0M over
arm : 134/900M (cur/typical) 0M over
auth : 226/655M (cur/typical) 0M over
ble_relay : 5/40M (cur/typical) 0M over
cfgm : 38/100M (cur/typical) 0M over
dbsync : 3/12M (cur/typical) 0M over
dds : 76/100M (cur/typical) 0M over
fpapps : 27/155M (cur/typical) 0M over
fw_visibility : 375/350M (cur/typical) 25M over
ha_mgr : 53/60M (cur/typical) 0M over
isakmpd : 7/450M (cur/typical) 0M over
l2tpd : 3/12M (cur/typical) 0M over
mdns : 500/100M (cur/typical) 400M over
mobileip : 4/12M (cur/typical) 0M over
ofa : 4/350M (cur/typical) 0M over
ospf : 3/12M (cur/typical) 0M over
pim : 9/10M (cur/typical) 0M over
pptpd : 3/10M (cur/typical) 0M over
profmgr : 7/12M (cur/typical) 0M over
slb : 4/12M (cur/typical) 0M over
snmpd : 6/10M (cur/typical) 0M over
stm : 216/954M (cur/typical) 0M over
ucm : 22/120M (cur/typical) 0M over
udbserver : 4/10M (cur/typical) 0M over
web_cc : 5/90M (cur/typical) 0M over
wms : 107/400M (cur/typical) 0M over

Brian
Frequent Contributor I

Re: MDNS Memory over subscribed

these memory thresholds dont really have any way to adust to what is normal for your deployment - take the 400M over with a grain of salt initially.

 

The more important thing will be if the value is climbing which may be indicative of a memory leak. You can monitor the output of "show memory debug" and "show memory mdns" over a couple of days to see what its behavior is (similarly you can view the controller overall memory use if you have airwave, how is it behaving over 3 days or a week)

 

if this is just heavy mdns usage the value should be quite static, if the value is climbing each day and not receding when the busy peak of the day is over then you may have a leak (wont be the first one mdns has seen).

 

heavy mem usage in mdns is expected if you have a lot of mdns devices in the network, it probably shouldnt cause impact to end devices unless something is full or taking too long, the logs may give a clue to that.

 

first things first - is it stable or increasing monotonically? how does it behave over a longer term ?

 

 

Occasional Contributor II

Re: MDNS Memory over subscribed

Thank you for the reply...I've monitored it over the last couple of weeks. I see the mdns memory changing. Sometimes on some controllers its over 1 gigabit oversubscribed. Then on the next day its not over subscribed at all. I'm going to monitor it via Airwave and see how the history matches up.

Brian
Frequent Contributor I

Re: MDNS Memory over subscribed

the key thing to check then is if the mdns process has crashed when it peaked - if it hasnt then it may just be growing and shrinking based on the load. Check to see if it has restarted using show process monitor statistics (and then tar crash -> log a ticket if it has)

 

 

Occasional Contributor II

Re: MDNS Memory over subscribed

Okay thanks...Will do.

Brian
Occasional Contributor II

Re: MDNS Memory over subscribed

After running the command it looks like the mdns has restarted 21 times on this particular controller.

 

mswitch/bin/mdns             PROCESS_RUNNING  -                 21        240            3                Sat Dec  9 04:56:18 2017

 

I'm going to open a case with TAC....Thanks for the insight and help.

Brian
Search Airheads
cancel
Showing results for 
Search instead for 
Did you mean: