"Read failure - Connection closed by peer"

[gael.soude]

Hello, we just deployed a cluster of Stingray Trafic Manager in our company in version 9.3 64bits.

After a while we encounter some issues with a Web Service used for SSO authentication.

Many times a day this Web service through Stingray is very slow or doesn't even work and the only message we have in the logs is :

"Read failure - Connection closed by peer"

I have no idea what it is and why it happends... And I can't find any documentation explaining the error messages with Stingray.

Can you explain me please what could be the problem ?

[felixdpg]

Try increasing the connection time to the backend servers.

Regards

Felix

[markbod · Accepted solution]

That message is seen when you have either client and/or server connection error reporting enabled. The setting is under Services -> Virtual Servers -> VSName -> Connection Management -> Connection Error Settings

The "Read failure" message can be seen when either the client or server unexpectedly close the connection to Stingray. You should see lots more information after the error providing you with details of the connection which had the problem. The first character should be either an 'S' or a 'C' to indicate whether it was a lient or erver connection problem. Eg:

If you get the error on the same url for a connection to the server, then it might indicate that a particular application or script is failing.

Cheers,

Mark

[gael.soude]

Hello, thanks a lot for the explanations.

As all the 'Read failure' events are on the Client side I changed the timeout settings from 10s to 20s on the Virtual Server. I think that what is hapenning is that the webservice is very slow to process and that 10s was not enought.

The client sent a Web request for authentication and sometimes it can take more than 10s before recieving any answer from the server !

[gael.soude]

Regarding this same problem

I'm wondering how the 'Client side' keep alive on the Virtual server works...

How can I control when the timeout is send and at what interval ?

Here is my problem :

We have a client connecting to a webservice in HTTPS.

On the client the Web connection timeout is hardcoded to 10 seconds (can't change it)

The problem is that sometimes the webservice can take more than 10s to answer.

My idea was to setup a Keep Alive on the Stealhead in order to maintain the client connection more than 10s

But i don't know what parameter to change and how.

I think this would be something like keepalive_timeout but I don't understand this parameter well.

Can you please clarify me on this point