Hey guys! Just wanted to share a little tip on how to increase Observium polling performance on juniper boxes. This might be old news for some but I just found it. You can create a filter on junos-device that filters out certain interfaces from SNMP view. https://www.juniper.net/documentation/en_US/junos/topics/task/configuration/...
I can recommend using: set snmp filter-interfaces all-internal-interfaces set snmp filter-interfaces interfaces "^sxe-.*"
This will filter out all useless internal interfaces from polling and discovery by Observium. Especially when you have large juniper switch stacks the total number of internal interfaces are a lot which slows down polling. Filtering them out actually makes a big difference.
If you got a lot of juniper gear, try it out and please share the speedup in pollingtime!
/Markus
What was your speedup on polling time? 🤭
Sent from BlueMail
On 17 Dec 2020, 20:43, at 20:43, Markus Klock via observium observium@observium.org wrote:
Hey guys! Just wanted to share a little tip on how to increase Observium polling performance on juniper boxes. This might be old news for some but I just found it. You can create a filter on junos-device that filters out certain interfaces from SNMP view. https://www.juniper.net/documentation/en_US/junos/topics/task/configuration/...
I can recommend using: set snmp filter-interfaces all-internal-interfaces set snmp filter-interfaces interfaces "^sxe-.*"
This will filter out all useless internal interfaces from polling and discovery by Observium. Especially when you have large juniper switch stacks the total number of internal interfaces are a lot which slows down polling. Filtering them out actually makes a big difference.
If you got a lot of juniper gear, try it out and please share the speedup in pollingtime!
/Markus
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
We include a few others, but it looks like we are missing the sxe-.* pattern which is present on some of our higher end gear.
set snmp filter-interfaces interfaces lsi.* set snmp filter-interfaces interfaces cbp.* set snmp filter-interfaces interfaces demux* set snmp filter-interfaces interfaces pime* set snmp filter-interfaces interfaces pimd* set snmp filter-interfaces interfaces pip* set snmp filter-interfaces interfaces tap* set snmp filter-interfaces interfaces lo0.16384 set snmp filter-interfaces interfaces "[g|x|a]e.*.32767" set snmp filter-interfaces interfaces "et.*.32767" set snmp filter-interfaces interfaces jsrv* set snmp filter-interfaces interfaces lo0.16385
________________________________ From: observium observium-bounces@observium.org on behalf of Markus Klock via observium observium@observium.org Sent: Thursday, December 17, 2020 12:43:19 PM To: Observium Network Observation System Cc: Markus Klock Subject: [Observium] Juniper polling performance tip
Hey guys! Just wanted to share a little tip on how to increase Observium polling performance on juniper boxes. This might be old news for some but I just found it. You can create a filter on junos-device that filters out certain interfaces from SNMP view. https://www.juniper.net/documentation/en_US/junos/topics/task/configuration/...
I can recommend using: set snmp filter-interfaces all-internal-interfaces set snmp filter-interfaces interfaces "^sxe-.*"
This will filter out all useless internal interfaces from polling and discovery by Observium. Especially when you have large juniper switch stacks the total number of internal interfaces are a lot which slows down polling. Filtering them out actually makes a big difference.
If you got a lot of juniper gear, try it out and please share the speedup in pollingtime!
/Markus
I've added this to the docs
https://docs.observium.org/device_juniper/
From: observium observium-bounces@observium.org On Behalf Of Milton Ngan via observium Sent: 17 December 2020 20:54 To: Observium observium@observium.org Cc: Milton Ngan milton@valvesoftware.com Subject: Re: [Observium] Juniper polling performance tip
We include a few others, but it looks like we are missing the sxe-.* pattern which is present on some of our higher end gear.
set snmp filter-interfaces interfaces lsi.*
set snmp filter-interfaces interfaces cbp.*
set snmp filter-interfaces interfaces demux*
set snmp filter-interfaces interfaces pime*
set snmp filter-interfaces interfaces pimd*
set snmp filter-interfaces interfaces pip*
set snmp filter-interfaces interfaces tap*
set snmp filter-interfaces interfaces lo0.16384
set snmp filter-interfaces interfaces "[g|x|a]e.*.32767"
set snmp filter-interfaces interfaces "et.*.32767"
set snmp filter-interfaces interfaces jsrv*
set snmp filter-interfaces interfaces lo0.16385
_____
From: observium <observium-bounces@observium.org mailto:observium-bounces@observium.org > on behalf of Markus Klock via observium <observium@observium.org mailto:observium@observium.org > Sent: Thursday, December 17, 2020 12:43:19 PM To: Observium Network Observation System Cc: Markus Klock Subject: [Observium] Juniper polling performance tip
Hey guys!
Just wanted to share a little tip on how to increase Observium polling performance on juniper boxes.
This might be old news for some but I just found it.
You can create a filter on junos-device that filters out certain interfaces from SNMP view.
https://www.juniper.net/documentation/en_US/junos/topics/task/configuration/ snmp-filter-interfaces-configuring-junos-nm.html
I can recommend using:
set snmp filter-interfaces all-internal-interfaces
set snmp filter-interfaces interfaces "^sxe-.*"
This will filter out all useless internal interfaces from polling and discovery by Observium.
Especially when you have large juniper switch stacks the total number of internal interfaces are a lot which slows down polling.
Filtering them out actually makes a big difference.
If you got a lot of juniper gear, try it out and please share the speedup in pollingtime!
/Markus
This is pretty handy thanks for the tip. I've always been in the habit of ignoring these interfaces when I saw them in the interface lists of snmp tools. all-internal-interfaces hid most.
I added: set snmp filter-interfaces interfaces lsi* set snmp filter-interfaces interfaces dsc* set snmp filter-interfaces interfaces tap* set snmp filter-interfaces interfaces pim* set snmp filter-interfaces interfaces mtun* set snmp filter-interfaces interfaces demux* set snmp filter-interfaces interfaces cbp* set snmp filter-interfaces interfaces pip* set snmp filter-interfaces interfaces pp* set snmp filter-interfaces interfaces "[g|x]e-.*.0$" set snmp filter-interfaces interfaces "ae[0-9]+.0$" set snmp filter-interfaces interfaces "et-.*.0$" set snmp filter-interfaces interfaces vtep* set snmp filter-interfaces interfaces esi* set snmp filter-interfaces interfaces jsrv* set snmp filter-interfaces interfaces rbeb* set snmp filter-interfaces interfaces gre set snmp filter-interfaces interfaces irb
Now my port list is super clean. I'm not entirely sure yet what problems I will cause by filtering out the .0 interfaces. In observium it looks like it doesn't learn the IP addresses of each interface.
MX480 here, snmpgets dropped 40-50 out of just over 100 and poller time dropped 5-6 seconds out of 67 average. The time saved seems mostly scrolling through shorter port lists now that the interface list is a fraction of the size. In observium many of these are covered by $bad_if lists. If you have multiple things doing snmp polling it might be better to do this in junos than maintain a $bad_if list for each polling program.
On Thu, Dec 17, 2020 at 1:06 PM Adam Armstrong via observium observium@observium.org wrote:
I’ve added this to the docs
https://docs.observium.org/device_juniper/
From: observium observium-bounces@observium.org On Behalf Of Milton Ngan via observium Sent: 17 December 2020 20:54 To: Observium observium@observium.org Cc: Milton Ngan milton@valvesoftware.com Subject: Re: [Observium] Juniper polling performance tip
We include a few others, but it looks like we are missing the sxe-.* pattern which is present on some of our higher end gear.
set snmp filter-interfaces interfaces lsi.*
set snmp filter-interfaces interfaces cbp.*
set snmp filter-interfaces interfaces demux*
set snmp filter-interfaces interfaces pime*
set snmp filter-interfaces interfaces pimd*
set snmp filter-interfaces interfaces pip*
set snmp filter-interfaces interfaces tap*
set snmp filter-interfaces interfaces lo0.16384
set snmp filter-interfaces interfaces "[g|x|a]e.*.32767"
set snmp filter-interfaces interfaces "et.*.32767"
set snmp filter-interfaces interfaces jsrv*
set snmp filter-interfaces interfaces lo0.16385
From: observium observium-bounces@observium.org on behalf of Markus Klock via observium observium@observium.org Sent: Thursday, December 17, 2020 12:43:19 PM To: Observium Network Observation System Cc: Markus Klock Subject: [Observium] Juniper polling performance tip
Hey guys!
Just wanted to share a little tip on how to increase Observium polling performance on juniper boxes.
This might be old news for some but I just found it.
You can create a filter on junos-device that filters out certain interfaces from SNMP view.
https://www.juniper.net/documentation/en_US/junos/topics/task/configuration/...
I can recommend using:
set snmp filter-interfaces all-internal-interfaces
set snmp filter-interfaces interfaces "^sxe-.*"
This will filter out all useless internal interfaces from polling and discovery by Observium.
Especially when you have large juniper switch stacks the total number of internal interfaces are a lot which slows down polling.
Filtering them out actually makes a big difference.
If you got a lot of juniper gear, try it out and please share the speedup in pollingtime!
/Markus
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
Beware of timing in cli, it actually takes a lot of processing to generate the tables, so it can add a significant amount of time.
When the poller is run wrapped, it doesn't generate the tables, and the times will be different.
Adam.
-----Original Message----- From: Chris Laffin chris.laffin@automattic.com Sent: 19 December 2020 00:39 To: Observium observium@observium.org Cc: Adam Armstrong adama@observium.org Subject: Re: [Observium] Juniper polling performance tip
This is pretty handy thanks for the tip. I've always been in the habit of ignoring these interfaces when I saw them in the interface lists of snmp tools. all-internal-interfaces hid most.
I added: set snmp filter-interfaces interfaces lsi* set snmp filter-interfaces interfaces dsc* set snmp filter-interfaces interfaces tap* set snmp filter-interfaces interfaces pim* set snmp filter-interfaces interfaces mtun* set snmp filter-interfaces interfaces demux* set snmp filter-interfaces interfaces cbp* set snmp filter-interfaces interfaces pip* set snmp filter-interfaces interfaces pp* set snmp filter-interfaces interfaces "[g|x]e-.*.0$" set snmp filter-interfaces interfaces "ae[0-9]+.0$" set snmp filter-interfaces interfaces "et-.*.0$" set snmp filter-interfaces interfaces vtep* set snmp filter-interfaces interfaces esi* set snmp filter-interfaces interfaces jsrv* set snmp filter-interfaces interfaces rbeb* set snmp filter-interfaces interfaces gre set snmp filter-interfaces interfaces irb
Now my port list is super clean. I'm not entirely sure yet what problems I will cause by filtering out the .0 interfaces. In observium it looks like it doesn't learn the IP addresses of each interface.
MX480 here, snmpgets dropped 40-50 out of just over 100 and poller time dropped 5-6 seconds out of 67 average. The time saved seems mostly scrolling through shorter port lists now that the interface list is a fraction of the size. In observium many of these are covered by $bad_if lists. If you have multiple things doing snmp polling it might be better to do this in junos than maintain a $bad_if list for each polling program.
On Thu, Dec 17, 2020 at 1:06 PM Adam Armstrong via observium observium@observium.org wrote:
I’ve added this to the docs
https://docs.observium.org/device_juniper/
From: observium observium-bounces@observium.org On Behalf Of Milton Ngan via observium Sent: 17 December 2020 20:54 To: Observium observium@observium.org Cc: Milton Ngan milton@valvesoftware.com Subject: Re: [Observium] Juniper polling performance tip
We include a few others, but it looks like we are missing the sxe-.* pattern which is present on some of our higher end gear.
set snmp filter-interfaces interfaces lsi.*
set snmp filter-interfaces interfaces cbp.*
set snmp filter-interfaces interfaces demux*
set snmp filter-interfaces interfaces pime*
set snmp filter-interfaces interfaces pimd*
set snmp filter-interfaces interfaces pip*
set snmp filter-interfaces interfaces tap*
set snmp filter-interfaces interfaces lo0.16384
set snmp filter-interfaces interfaces "[g|x|a]e.*.32767"
set snmp filter-interfaces interfaces "et.*.32767"
set snmp filter-interfaces interfaces jsrv*
set snmp filter-interfaces interfaces lo0.16385
From: observium observium-bounces@observium.org on behalf of Markus Klock via observium observium@observium.org Sent: Thursday, December 17, 2020 12:43:19 PM To: Observium Network Observation System Cc: Markus Klock Subject: [Observium] Juniper polling performance tip
Hey guys!
Just wanted to share a little tip on how to increase Observium polling performance on juniper boxes.
This might be old news for some but I just found it.
You can create a filter on junos-device that filters out certain interfaces from SNMP view.
https://www.juniper.net/documentation/en_US/junos/topics/task/configur ation/snmp-filter-interfaces-configuring-junos-nm.html
I can recommend using:
set snmp filter-interfaces all-internal-interfaces
set snmp filter-interfaces interfaces "^sxe-.*"
This will filter out all useless internal interfaces from polling and discovery by Observium.
Especially when you have large juniper switch stacks the total number of internal interfaces are a lot which slows down polling.
Filtering them out actually makes a big difference.
If you got a lot of juniper gear, try it out and please share the speedup in pollingtime!
/Markus
observium mailing list observium@observium.org http://postman.memetic.org/cgi-bin/mailman/listinfo/observium
participants (4)
-
Adam Armstrong
-
Chris Laffin
-
Markus Klock
-
Milton Ngan