Thread: BUG #14062: pg_dump dies after dumping first 60 gigabytes of text for large table

The following bug has been logged on the website:

Bug reference:      14062
Logged by:          Dan Liddell
Email address:      dll@sonic.net
PostgreSQL version: 9.4.6
Operating system:   Red Hat Enterprise Linux Server release 7.2 (Maipo
Description:


For an approximately 500 gigabyte database, pg_dump dies after approximately
60 gb of text output, with an error:

pg_dump: Dumping the contents of table "datum" failed: PQgetCopyData()
failed.
pg_dump: Error message from server: SSL SYSCALL error: EOF detected
pg_dump: The command was: COPY hrrr.datum (dr_id, qtime, grid_id, gust,
ugrd80, vgrd80, pres, tmp, dpt, ugrd10, vgrd10, dswrf) TO stdout;

The command used is:
 pg_dump -n hrrr --username=webuser --host=<some AWS RDS host> hrrr_db

The host is remote, and the connection is encrypted.

dll@sonic.net writes:
> For an approximately 500 gigabyte database, pg_dump dies after approximately
> 60 gb of text output, with an error:

> pg_dump: Dumping the contents of table "datum" failed: PQgetCopyData()
> failed.
> pg_dump: Error message from server: SSL SYSCALL error: EOF detected

Sounds like an SSL renegotiation bug.  We gave up on that mess and
disabled renegotiation by default as of 9.4.5 or so, but maybe you have
an explicit nonzero setting for ssl_renegotiation_limit?

            regards, tom lane

Amazon RDS instance has a default value of 0 for ssl_renegotiation_limit for
the postgres 9.4 default parameters, according to the RDS Dashboard.

However, when we disable SSL connections the problem evaporates. Good enough
for me.

Thanks, Tom!

On 4/4/2016 12:56 PM, Tom Lane wrote:
> dll@sonic.net writes:
>> For an approximately 500 gigabyte database, pg_dump dies after approximately
>> 60 gb of text output, with an error:
>
>> pg_dump: Dumping the contents of table "datum" failed: PQgetCopyData()
>> failed.
>> pg_dump: Error message from server: SSL SYSCALL error: EOF detected
>
> Sounds like an SSL renegotiation bug.  We gave up on that mess and
> disabled renegotiation by default as of 9.4.5 or so, but maybe you have
> an explicit nonzero setting for ssl_renegotiation_limit?
>
>             regards, tom lane
>

dll wrote
> Amazon RDS instance has a default value of 0 for ssl_renegotiation_limit
> for
> the postgres 9.4 default parameters, according to the RDS Dashboard.
> 
> However, when we disable SSL connections the problem evaporates. Good
> enough
> for me.
> 
> Thanks, Tom!
> 
> On 4/4/2016 12:56 PM, Tom Lane wrote:
>> 

> dll@

>  writes:
>>> For an approximately 500 gigabyte database, pg_dump dies after
>>> approximately
>>> 60 gb of text output, with an error:
>> 
>>> pg_dump: Dumping the contents of table "datum" failed: PQgetCopyData()
>>> failed.
>>> pg_dump: Error message from server: SSL SYSCALL error: EOF detected
>> 
>> Sounds like an SSL renegotiation bug.  We gave up on that mess and
>> disabled renegotiation by default as of 9.4.5 or so, but maybe you have
>> an explicit nonzero setting for ssl_renegotiation_limit?
>> 
>>             regards, tom lane
>> 
> 
> Hi Tom,
> 
> In my case, ssl disable. 
> 
> it's table "tbwww", size around 70GB. My config server it's similar Redhat
> 7.2 and Postgres 9.5.3
> 
> For erro:
> Line 21172: pg_dump: Dumping the contents of table "tbwww" failed:
> PQgetCopyData() failed.
> Line 21173: pg_dump: Error message from server: server closed the
> connection unexpectedly
> Line 21176: pg_dump: The command was: COPY file.tbwww (fileid, filename,
> filenamedownload, file, filedate, coduser) TO stdout;
> 
> -- 
> Sent via pgsql-bugs mailing list (

> pgsql-bugs@

> )
> To make changes to your subscription:
> http://www.postgresql.org/mailpref/pgsql-bugs





--
Sent from: http://www.postgresql-archive.org/PostgreSQL-bugs-f2117394.html