Thoughts on Internet Connection Monitoring?

[Steve_Pearson]

I was wondering what everyones thoughts were on Internet Connection monitoring?

I'm configuring a handful of Spark 1535/55 boxes for branch offices, each has a single leased line. I'm wondering if I should leave the connection monitoring on or turn it off.

The reason I'm thinking about this is down to recent issues on a Spark 1600 that kept dropping the internet connection, due to what it considered to be excessive packet loss, (although no testing could verify this), and as it was a single internet connection it would obviously bring the connectivity down totally causing outage on a production site. TAC recommended to disable the connection monitoring completely on this box as it only has one connection, so I'm wondering about the 1500's too!

[AkosBakos]

Hi @Steve_Pearson 

Lets say, leased line is stable. 

Because of you have a couple of same SPARKs, what if you enable the monitoring only on one SPARK then will see what happening?

Akos

I seems, the answer will be in testing...

[AkosBakos]

And one more:

Do we know the exessive packet loss is still persists? 

Can we exclude HW (and cable) problems?

Ákos

[Steve_Pearson]

On the 1600, personally, I think the issue is related to the SDWAN blade (even with it switched off!) as when I ran a test that had the WAN port connected to a little switch, a test laptop connected to the same switch and the connection monitoring using the IP of the laptop to ping and it still reported packet loss. never had a problem on earlier firmware!

At the moment i'm just trying to see what people normally do with connection monitoring on the Spark 1500 devices

[the_rock]

I saw another post you had about these devices being centrally managed, but does not seem you can enable monitoring blade on the smart console object for these devices. That would have been another option.

Andy

[Amir_Ayalon · Accepted solution]

Hi Guys,

The Connection monitoring on the small boxes and on the 1600/1800/2000 is the same.

Please notice that enabling it when you have only one internet connection, is for statistics and graph only, so if you don't need these, you can turn it off.

Also important to mention, that sometimes you need to tweak the setting for it to properly work. increase the latency, increase Jitter. this is as a result of different line condition, different type on internet connectivity, and more (important especially with satellite connectivity, or LTE/5G, but recommended to all to review their setting)

Soon we will release a R81.10.15 Jumbo replacement with some probing fixes.

in the meanwhile, if you are working with Cluster, or have one internet connection - you can disable internet monitoring. 

[perfect4situa]

I think this is not the solution because you should be able to monitor connection quality even with only one internet isp.

Furthermore I need to enable monitoring to make SD-WAN work in other cases where I've got more than one isp.

Can someone just explain why does monitoring take this result, for example in the test with pinging directly the laptop I can't expect to receive a failure probing...

I'd just like to know if this is a recognised problem and if there's any solution to make monitoring and SD-WAN work in the meanwhile...

Thank you

[the_rock]

Did TAC end up providing you any options for it?

Andy

[StevePearson]

Hi Andy,

In short, no!

The solution was to not monitor the connection, however there have been some huge issues with this customer. Without connection monitoring, I was able to get the cluster to run for a while but with hundreds of errors relating to vlans on some ports. (apparently vlans are not supported in a cluster at the moment!) Then the cluster would simply stop passing traffic for no reason, with the only fix being a hard reboot. Removing the vlans did not help the issue but caused yet more trouble for the customer! Basically we were suggesting to TAC that the clustering was not fit for purpose and we ended up on remote sessions with R&D, special fixes being supplied and kernel parameter changes etc, just to try to stabilise it! (It is now stable "ish" but still without vlans)

What it comes down to are some major issues (their words not ours!) in the R81.10.15 firmware and we should wait for R81.10.20 is resolve all of these issues!

[the_rock]

Yea, I can totally see where customer is coming from, definitely not ideal situation : - (

Andy