Re: Prevent loss of query text when pgAdmin crashes or freezes

I often have multiple SQL query windows open in pgAdmin (v1.14.3 on
Windows 7 x64). From time to time pgAdmin will crash or freeze (become
unresponsive) and would have to be killed. When this happens all unsaved
query text is lost, which is extremely frustrating. It's by far the
biggest problem for me in using pgAdmin.

Please implement some way to recover the query text from such crashes.
Ideally something like what Firefox or Chrome do: when restarted after a
crash they offer the option to restore previous windows (including form
data, at least in Chrome). This would require that pgAdmin saves query
text in the background on a regular basis (at least before running any
query).

Another great way to mitigate this would be to run each SQL query window
in a separate process, so it could crash without bringing down the rest.
However, the main pgAdmin process could crash, too, and this should not
kill the query window processes without allowing data to be saved.

Regards,

Evan

On Mon, 2012-07-02 at 12:32 +1000, Evan Martin wrote:
> I often have multiple SQL query windows open in pgAdmin (v1.14.3 on
> Windows 7 x64). From time to time pgAdmin will crash or freeze (become
> unresponsive) and would have to be killed. When this happens all unsaved
> query text is lost, which is extremely frustrating. It's by far the
> biggest problem for me in using pgAdmin.
>
> Please implement some way to recover the query text from such crashes.
> Ideally something like what Firefox or Chrome do: when restarted after a
> crash they offer the option to restore previous windows (including form
> data, at least in Chrome). This would require that pgAdmin saves query
> text in the background on a regular basis (at least before running any
> query).
>
> Another great way to mitigate this would be to run each SQL query window
> in a separate process, so it could crash without bringing down the rest.
> However, the main pgAdmin process could crash, too, and this should not
> kill the query window processes without allowing data to be saved.
>

I though we already had this in our TODO list, but I can't find it. So I
added it.

Thanks for the idea.

--
Guillaume
http://blog.guillaume.lelarge.info
http://www.dalibo.com

On Mon, Jul 2, 2012 at 9:44 PM, Guillaume Lelarge
<guillaume(at)lelarge(dot)info> wrote:
> On Mon, 2012-07-02 at 12:32 +1000, Evan Martin wrote:
>> I often have multiple SQL query windows open in pgAdmin (v1.14.3 on
>> Windows 7 x64). From time to time pgAdmin will crash or freeze (become
>> unresponsive) and would have to be killed. When this happens all unsaved
>> query text is lost, which is extremely frustrating. It's by far the
>> biggest problem for me in using pgAdmin.
>>
>> Please implement some way to recover the query text from such crashes.
>> Ideally something like what Firefox or Chrome do: when restarted after a
>> crash they offer the option to restore previous windows (including form
>> data, at least in Chrome). This would require that pgAdmin saves query
>> text in the background on a regular basis (at least before running any
>> query).
>>
>> Another great way to mitigate this would be to run each SQL query window
>> in a separate process, so it could crash without bringing down the rest.
>> However, the main pgAdmin process could crash, too, and this should not
>> kill the query window processes without allowing data to be saved.
>>
>
> I though we already had this in our TODO list, but I can't find it. So I
> added it.

I'd far rather see the bug fixed, but without any details that seems
tricky. FWIW, I use pgAdmin on Windows and Mac daily, and don't run
into crashes or hangs.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

pgAdmin crashes from time to time when connection is lost. For example
when using notebook, with pgAdmin, you often change networks, hibernate,
suspend, using direct connection or VPN etc
Such changes break connections to databases. I found that closing
pgAdmin is quicker solution than reconnecting (see bellow). But closing
it while in this state, often causes pgAdmin crash.

About reconnection: I'm curious when this feature will be finally
finished. Some, let's say, prototype has been implemented some time ago,
but using it is annoying. After disconnection, you have to confirm five
times the need of reconnection when using object browser. When using
query window, you have run command twice: after asking for reconnection,
command end up with error; second run finally works.
Currently if I'm connected to multiple databases and lost connection,
killing pgadmin is fastest way to reconnect ;)
> I'd far rather see the bug fixed, but without any details that seems
> tricky. FWIW, I use pgAdmin on Windows and Mac daily, and don't run
> into crashes or hangs.
>
>

Fixing the bug and implementing this are not mutually exclusive. Of
course, it should be fixed, but that's just one bug. I've experienced at
least 3 different kinds of crashes. This is about mitigating the impact
of all crashing bugs, which is important, because they will never be
eliminated completely.

Thanks for adding it to the TODO list, Guillaume. I'll post separately
about this particular crash if I can reproduce it reliably.

Regards,
Evan

On 3/07/2012 6:28 PM, Dave Page wrote:
> On Mon, Jul 2, 2012 at 9:44 PM, Guillaume Lelarge
> <guillaume(at)lelarge(dot)info> wrote:
>>
>> I though we already had this in our TODO list, but I can't find it. So I
>> added it.
> I'd far rather see the bug fixed, but without any details that seems
> tricky. FWIW, I use pgAdmin on Windows and Mac daily, and don't run
> into crashes or hangs.
>

On Tue, Jul 3, 2012 at 9:58 AM, Michal Kozusznik
<kozusznik(dot)michal(at)ifortuna(dot)cz> wrote:
> pgAdmin crashes from time to time when connection is lost. For example when
> using notebook, with pgAdmin, you often change networks, hibernate, suspend,
> using direct connection or VPN etc
> Such changes break connections to databases. I found that closing pgAdmin is
> quicker solution than reconnecting (see bellow). But closing it while in
> this state, often causes pgAdmin crash.
>
> About reconnection: I'm curious when this feature will be finally finished.
> Some, let's say, prototype has been implemented some time ago, but using it
> is annoying. After disconnection, you have to confirm five times the need of
> reconnection when using object browser. When using query window, you have
> run command twice: after asking for reconnection, command end up with error;
> second run finally works.

Rewriting the connection management code to deal with this is far from
a trivial task. It's dependent on someone finding the time to do it.

--
Dave Page
Blog: http://pgsnake.blogspot.com
Twitter: @pgsnake

EnterpriseDB UK: http://www.enterprisedb.com
The Enterprise PostgreSQL Company

On Tue, Jul 3, 2012 at 12:55 PM, Evan Martin
<postgresql(at)realityexists(dot)net>wrote:

> Fixing the bug and implementing this are not mutually exclusive. Of
> course, it should be fixed, but that's just one bug. I've experienced at
> least 3 different kinds of crashes. This is about mitigating the impact of
> all crashing bugs, which is important, because they will never be
> eliminated completely.

Especially if the crash has nothing to do with pgAdmin (like Windows itself
crashing).

Please, don't start discussion about windows against linux stability.
I'm working on W7 and I really don't remember if it crashed even once.
I have a collogue who sit next to me, and he also reports pgadmin crash
often.
We suspect that pgadmin has problems with manage disconnections.

Dne 3.7.2012 15:52, Vik Reykja napsal(a):
> On Tue, Jul 3, 2012 at 12:55 PM, Evan Martin
> <postgresql(at)realityexists(dot)net <mailto:postgresql(at)realityexists(dot)net>>
> wrote:
>
> Fixing the bug and implementing this are not mutually exclusive.
> Of course, it should be fixed, but that's just one bug. I've
> experienced at least 3 different kinds of crashes. This is about
> mitigating the impact of all crashing bugs, which is important,
> because they will never be eliminated completely.
>
>
> Especially if the crash has nothing to do with pgAdmin (like Windows
> itself crashing).

I meant a colleague working with Linux

Dne 3.7.2012 16:22, Michal Kozusznik napsal(a):
> Please, don't start discussion about windows against linux stability.
> I'm working on W7 and I really don't remember if it crashed even once.
> I have a collogue who sit next to me, and he also reports pgadmin
> crash often.
> We suspect that pgadmin has problems with manage disconnections.
>
> Dne 3.7.2012 15:52, Vik Reykja napsal(a):
>> On Tue, Jul 3, 2012 at 12:55 PM, Evan Martin
>> <postgresql(at)realityexists(dot)net <mailto:postgresql(at)realityexists(dot)net>>
>> wrote:
>>
>> Fixing the bug and implementing this are not mutually exclusive.
>> Of course, it should be fixed, but that's just one bug. I've
>> experienced at least 3 different kinds of crashes. This is about
>> mitigating the impact of all crashing bugs, which is important,
>> because they will never be eliminated completely.
>>
>>
>> Especially if the crash has nothing to do with pgAdmin (like Windows
>> itself crashing).
>
>
>

On Tue, Jul 3, 2012 at 4:22 PM, Michal Kozusznik <
kozusznik(dot)michal(at)ifortuna(dot)cz> wrote:

> Please, don't start discussion about windows against linux stability.
> I'm working on W7 and I really don't remember if it crashed even once.
>

Please don't top post.

I don't believe I said anything about linux. You reported using Windows, I
also use Windows, it sometimes crashes on me and I lose my work. That's
somewhat my own fault for not saving often enough, but if pgAdmin knew how
to recover for me, I'd be happy.

This is not a flame war.

> I have a collogue who sit next to me, and he also reports pgadmin crash
> often.
> We suspect that pgadmin has problems with manage disconnections.
>

It certainly does, but that is only one of the ways for a crash.

>
> Dne 3.7.2012 15:52, Vik Reykja napsal(a):
>
> On Tue, Jul 3, 2012 at 12:55 PM, Evan Martin <postgresql(at)realityexists(dot)net
> > wrote:
>
>> Fixing the bug and implementing this are not mutually exclusive. Of
>> course, it should be fixed, but that's just one bug. I've experienced at
>> least 3 different kinds of crashes. This is about mitigating the impact of
>> all crashing bugs, which is important, because they will never be
>> eliminated completely.
>
>
> Especially if the crash has nothing to do with pgAdmin (like Windows
> itself crashing).
>
>
>
>
>