Re: Looking for ideas on blocked sessions updating partitioned table with CLOB (solved)
From: Mladen Gogala <gogala.mladen_at_gmail.com>
Date: Wed, 19 Sep 2018 10:52:41 -0400
Message-ID: <7ebaf5a8-f1b4-b601-d964-495f327cd759_at_gmail.com>
Date: Wed, 19 Sep 2018 10:52:41 -0400
Message-ID: <7ebaf5a8-f1b4-b601-d964-495f327cd759_at_gmail.com>
Yes, that is the problem with selective acknowledgement or "SACK". To make the matter worse, your packet time stamps were turned off, so there was no protection against out of sequence packets. If the client expected acknowledgement for the TCP packet which would never come because selective acknowledgements were turned off, that would appear as an infinite wait. Here is a decent discussion of that parameter:
https://serverfault.com/questions/10955/when-to-turn-tcp-sack-off
On 09/18/2018 10:22 AM, Chris Taylor wrote:
>
> Running strace on the sessions on both the db server and the php
> server showed the sessions just sitting in either read() or write() to
> the network socket and never completing.
>
-- Mladen Gogala Database Consultant Tel: (347) 321-1217 -- http://www.freelists.org/webpage/oracle-lReceived on Wed Sep 19 2018 - 16:52:41 CEST