Re: Keep-alive?

Dave Page wrote:

>
>
>
>
>>-----Original Message-----
>>From: Adam H. Pendleton [mailto:fmonkey(at)fmonkey(dot)net]
>>Sent: 09 October 2003 13:31
>>To: Dave Page
>>Cc: Michiel_Lange(at)actuera(dot)nl; mark(at)nsa(dot)co(dot)il;
>>pgadmin-support(at)postgresql(dot)org
>>Subject: Re: [pgadmin-support] Keep-alive?
>>
>>Dave Page wrote:
>>
>>
>>
>>>2) Implement SQL level keep alives. This could a be simple SELECT
>>>version() query that is executed automatically after n seconds of
>>>inactivity.
>>>
>>>
>>>
>>>
>>>
>>This is what I had in mind, an application (i.e., layer 7)
>>level keep-alive.
>>
>>
>
>OK, just fiddling around here - attached is a patch that implements a
>heartbeat on the master connection to each server. Currently it beats
>every 5 seconds, though this would be made configurable/disable-able.
>
>It doesn't actually do much other than throw an error message if it gets
>an unexpected response.
>
>Thoughts, comments?
>
Yes, and a harsh one :-(

This is just what I meant that should *not* be implemented.
- It will throw an error every 5 seconds, after blocking the connection
for a failed heartbeat. Implemented this way, it's more a watchdog, that
won't stop barking.
- If a broken conn is detected, it must be marked dead.
- This feature is exotic, and must not be active by default.
- 5 seconds is ridiculously short.
- it should fire after n seconds of inactivity, not every n seconds
- this works only for the main window, not for those getting their own
connection. The wxWin timer will conflict with the threaded executed
queries, because it's not aware of previously or currently executed queries.
- Does it work at all? pgConn is no window, so where should the message
come from?

Regards,
Andrea