Thread: streaming replication: could not receive data from client: Connection reset by peer
streaming replication: could not receive data from client: Connection reset by peer
From
Anson Abraham
Date:
I'm on Debian Wheezy running postgres 9.3
both boxes are identical.
I see in log file on slaves:
LOG: could not receive data from client: Connection reset by peer
OpenSSL is the same version on master and slaves. The libssl is also too.
I set ssl_renegotiation=0. So not sure why i'm seeing this in the logs on my slave environments. doing research (aka googling), i see that everyone says to set ssl_renegotiation=0 but already done that. also mentions of libssl library / OpenSSL could be different versions, but that's not the case either.
Any ideas what else I should try doing?
-Anson
Re: streaming replication: could not receive data from client: Connection reset by peer
From
Albe Laurenz
Date:
Anson Abraham wrote: > I'm on Debian Wheezy running postgres 9.3 > > both boxes are identical. > > I see in log file on slaves: > LOG: could not receive data from client: Connection reset by peer That looks to me like a client that is connected to the slave is dying. Do you have hot standby turned on? Are there any other messages on master or slave around this time? Does this message come right away whan you start the slave or only after a certain time? > OpenSSL is the same version on master and slaves. The libssl is also too. Do you use SSL for streaming replication? What connection options does the slave use to connect to the master? > I set ssl_renegotiation=0. So not sure why i'm seeing this in the logs on my slave environments. > doing research (aka googling), i see that everyone says to set ssl_renegotiation=0 but already done > that. also mentions of libssl library / OpenSSL could be different versions, but that's not the case > either. That would help if you have a renegotiation problem, i.e. streaming replication has a problem after a time of correct operation. Yours, Laurenz Albe
Re: streaming replication: could not receive data from client: Connection reset by peer
From
Anson Abraham
Date:
No client connecting to the slave. It's just streamed replication for HA. This occurs when the slave starts immediately. SSL is used. And as I mentioned the libraries are identical on both slave and master. Interestingly, another slave that replicates from master does not have this issue. The one difference between the two slaves, the slave w/ the message is a VM. But other master dbs i have w/ slaves (that are VMs) and have no issues. So it's kind of odd, that this message occurs.
On Mon, Oct 21, 2013 at 6:50 AM, Albe Laurenz <laurenz.albe@wien.gv.at> wrote:
Anson Abraham wrote:That looks to me like a client that is connected to the slave
> I'm on Debian Wheezy running postgres 9.3
>
> both boxes are identical.
>
> I see in log file on slaves:
> LOG: could not receive data from client: Connection reset by peer
is dying.
Do you have hot standby turned on?
Are there any other messages on master or slave around this time?
Does this message come right away whan you start the slave or only
after a certain time?Do you use SSL for streaming replication?
> OpenSSL is the same version on master and slaves. The libssl is also too.
What connection options does the slave use to connect to the master?That would help if you have a renegotiation problem, i.e.
> I set ssl_renegotiation=0. So not sure why i'm seeing this in the logs on my slave environments.
> doing research (aka googling), i see that everyone says to set ssl_renegotiation=0 but already done
> that. also mentions of libssl library / OpenSSL could be different versions, but that's not the case
> either.
streaming replication has a problem after a time of correct operation.
Yours,
Laurenz Albe
Re: streaming replication: could not receive data from client: Connection reset by peer
From
Albe Laurenz
Date:
Anson Abraham wrote: > No client connecting to the slave. It's just streamed replication for HA. This occurs when the slave > starts immediately. SSL is used. And as I mentioned the libraries are identical on both slave and > master. Interestingly, another slave that replicates from master does not have this issue. The one > difference between the two slaves, the slave w/ the message is a VM. But other master dbs i have w/ > slaves (that are VMs) and have no issues. So it's kind of odd, that this message occurs. So, right when the slave starts, you get the message in the log, right? Is there anything else in the log? Is there anything in the master's log around that time? Try to set log_connections=on and log_disconnections=on, then you should at least see when the slave connects. Could there be a firewall or something that filters or disconnects sessions? Yours, Laurenz Albe