vSZ-H is going down beacause ElasticSearch is offline"

Question

Hi.

We have met problem which vSZ-H is going down beacause Elasticserch go down.

vSZ-H version is 3.5.1.x.

Beacuse Our customer serve only web-auth using wispr for public wifi service, all wlan service  is down now.

Before I have met same issue at other site, at that time I had resolved same problem using a command as "force recover escluster"

But the command doesn't find on vSZ-H.

The SZ resolved was SZ100.

Where is the command on vSZ-H.

And Why happens it.

Plz let me know about it.

unknown · Answer

Morning,

We have exactly the same problem and having spent some time with Ruckus support, it appears the issue is related to latency between the cluster members. According to the below T&C's, latency between clusters cannot be more than 8ms:
https://support.ruckuswireless.com/software/963-vsz-3-4-ga-software-release-ximg-image/tc

Apart from the above, there is absolutely no documentation that states that your clusters cannot have more than 8ms latency between them.

So the latency is the cause for ElasticSearch going down. So not only is that a problem, but you are unable to restart the ES service yourself. In order to restart ES, you need to contact support. I find this incredible!

Our vSZ-H devices are in geographically separate locations meaning that 8ms latency is near impossible! I have been informed that:

"Please do note that this behavior will change in future
releases where latency will not affect ES services."

No comment on WHEN these future releases will be made available. This means that we have purchased an extra licence that we cannot use.

michael_brado · Answer

Hello Medi, 
   Actually, you MUST power both Cluster devices at the SAME TIME (like within seconds of each other) in order for the Cassandra databases to synchronize correctly.

unknown · Answer

Hi.

I have faced same problem twice.

It's seriously bug or problem.

First case is standalone SZ100, second case is vSZ cluster.

If you are facing problem now, you can try this command as "force recover escluster".

In first case, I have solved it using this command.

Thanks.