Re: [sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2021-01-20 Thread Mike Bayer
On Wed, Jan 20, 2021, at 5:42 AM, Harry wrote: > With apologies for resuscitating an old thread, I came across this issue last > week, a 15-minute hang while an amazon RDS instance was doing a failover (we > are on SQLA 1.3.22). I need to do some more work on repro-ing the issue but: > >

Re: [sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2021-01-20 Thread Harry
setting socket options like this might be more of a psycopg2 thing? some more investigations here: https://stackoverflow.com/questions/26741175/psycopg2-db-connection-hangs-on-lost-network-connection On Wednesday, 20 January 2021 at 10:42:35 UTC Harry wrote: > With apologies for resuscitating

Re: [sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2021-01-20 Thread Harry
With apologies for resuscitating an old thread, I came across this issue last week, a 15-minute hang while an amazon RDS instance was doing a failover (we are on SQLA 1.3.22). I need to do some more work on repro-ing the issue but: This post on SO from last month

Re: [sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2016-02-07 Thread Joshua Ma
Forgot to send an update myself. We actually figured it out: I found that it was due to SQLA not properly detecting a failed connection - we get "SSL SYSCALL error" and the connection pool doesn't reopen + stays borked. Looks like SQLA has a few hard-coded string checks that it looks for and

Re: [sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2016-02-06 Thread Jonas Oscarsson
Hi Josh, Sorry for the late reply. Unfortunately we haven't had time/prioritized to look into it further. For the very few times when we've had to use the failover functionality, it has recovered quickly enough (within two minutes or so). Best, Jonas On Mon, Dec 21, 2015 at 9:19 PM Joshua Ma

Re: [sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2015-12-21 Thread Joshua Ma
Hi Jonas, Just wondering if you were able to characterize the behavior and had any ideas for gracefully handling RDS failover. From what I'm seeing, RDS will not cause errors, but will simply hang, so SQLA doesn't hit its "optimistic disconnect handling"

Re: [sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2014-10-18 Thread Jonas Oscarsson
Thanks for the tip, we haven't tried that yet although I think we would have the same behavior as it seems to be the next connection after the database is down that hangs (no matter if it's a SELECT 1 or an actual query). On Fri, Oct 17, 2014 at 5:38 PM, Jonathan Vanasco jvana...@gmail.com wrote:

[sqlalchemy] Re: Multi-AZ RDS on AWS, failover and SQLAlchemy

2014-10-17 Thread Jonathan Vanasco
Have you tried a Pessimistic Disconnect strategy? http://docs.sqlalchemy.org/en/latest/core/pooling.html#dealing-with-disconnects -- You received this message because you are subscribed to the Google Groups sqlalchemy group. To unsubscribe from this group and stop receiving emails from it,