need Keep-Alive...queries killed at 60 minutes?

[mattberns]

We are using vertica-python via dbt-vertica and we're finding that our log running queries are getting killed at 3600 seconds. We are told that we need to implement Keep-Alive since there are no packets being transacted for > 60 minutes. TCP connections that are idle are being killed at 60 minutes.

Is this a setting we can enable here? it appears that the way this is being done in some tools is to submit a query every N seconds over an open connection like SELECT 1. Does the driver support this?

[sitingren]

Hi @mattberns
There are related discussion at #205.

Basically, vertica-python provides connection_timeout as a connection parameter to control socket operations timeout (see README), queries can be affected by this. Please note that dbt-vertica is a third-party library that we don't support. If you want to set connection_timeout, you may need to change your libraries source code. The 3600 seconds comes from dbt-vertica:
https://github.com/mpcarter/dbt-vertica/search?q=timeout&type=

[mattberns]

@sitingren Thanks for this. Going to close this one. It turns out there is a timeout parameter over at dbt-vertica and this was set to connection-timeout in our repo which was misleading us ( because this setting passes through to connection-timeout This does seem to differ from the broken pipe issue which a colleague was experiencing less than the setting value, which we believe is due to a misconfiguration of the server.