reconnet problem and fix about nats.go HOT 10 CLOSED

Thanks for the report! That looks like it could be an issue, and I think it is the source of a bug we had internally found recently, but hadn't investigated deeper.

I don't know if @derekcollison will have the bandwidth to look at it in the next day or two, but I'll work on getting a test together to confirm it and see. I'll follow up once I can confirm it.

from nats.go.

The readLoop is recycled on each reconnect, meaning the Go routine for an existing connection is torn down and a new one created. I will look into whether or not there is a race with the state of the connection and a new readLoop upon reconnect. Thanks.

from nats.go.

looking forward for your reply.

Date: Wed, 17 Sep 2014 06:31:14 -0700
From: [email protected]
To: [email protected]
CC: [email protected]
Subject: Re: [nats] reconnet problem and fix (#29)

The readLoop is recycled on each reconnect, meaning the Go routine for an existing connection is torn down and a new one created. I will look into whether or not there is a race with the state of the connection and a new readLoop upon reconnect. Thanks.

¡ª
Reply to this email directly or view it on GitHub.

from nats.go.

How were you measuring the subsequent number of client connections? Looking at gnatsd or querying the clients themselves?

from nats.go.

Derekcollison,
First thanks for your reply. I have a 3 servers cluster. I use netstat -na | grep 4244 on each server to check how many clients connected to the server. Once I killed gnatsd I run the command to check which server got clients and plus all exist clients to check if any client is lost. After I killed and checked I start gnatsd on the server again, and I killed another server like previous.

from nats.go.

How long do you wait to restart the gnatsd server?

So to clarify, during this test you will see less clients successfully
reconnect to the server? Are all clients running in the same program? Is it
20 clients? Are they all Go based?

On Tue, Oct 7, 2014 at 6:17 AM, qinguoan [email protected] wrote:

Derekcollison,
First thanks for your reply. I have a 3 servers cluster. I use netstat
-na | grep 4244 on each server to check how many clients connected to the
server. Once I killed gnatsd I run the command to check which server got
clients and plus all exist clients to check if any client is lost. After I
killed and checked I start gnatsd on the server again, and I killed another
server like all what I did above.

—
Reply to this email directly or view it on GitHub
#29 (comment).

from nats.go.

I have located the issue, will work on a solution to the problem. Thanks again.

from nats.go.

sorry for my delay.

1. I wait for like a minute or a bit more.
2. Some will success but some failed.
3. Run on different machines.
4. All go based, use your lib.

from nats.go.

I have found the problem, working through a solution, should be posted this
week. Thanks.

On Wed, Oct 15, 2014 at 3:14 AM, qinguoan [email protected] wrote:

sorry for my delay.

1. I wait for like a minute or a bit more.
2. Some will success but some failed.
3. Run on different machines.
4. All go based, use your lib.

—
Reply to this email directly or view it on GitHub
#29 (comment).

from nats.go.

I have just fix this issue in my project, with derek‘s kindly help.
Thank all you guys~

from nats.go.