Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

FF 10 shutdown

42 views
Skip to first unread message

Dave Saville

unread,
Sep 30, 2012, 2:34:47 PM9/30/12
to
Takes tens of seconds with the hard drive going like the clappers. WPS
totally un responsive until it ends.

What the heck is it doing?
--
Regards
Dave Saville

Lewis Rosenthal

unread,
Sep 30, 2012, 3:56:42 PM9/30/12
to
Hey, Dave!

On 09/30/12 02:34 pm, Dave Saville thus wrote :
> Takes tens of seconds with the hard drive going like the clappers. WPS
> totally un responsive until it ends.
>
> What the heck is it doing?

Sounds like one or more of the sqlite databases needs some maintenance.

We had a good discussion about this in this ng back in June. Check that
thread for some tips (subject: SQLite vacuuming).

Also, Steve has some tips on this on his Warpzilla page (look for
SQLite): http://www.os2bbs.com/os2news/warpzilla.html

Also, you might want to glance at this:

https://developer.mozilla.org/en-US/docs/Storage/Performance

See if you have a particular extension which is cleaning up after itself
or otherwise involved with one of the databases (particularly
places.sqlite).

Best I can think of offhand. GL

Cheers/2

--
Lewis
-------------------------------------------------------------
Lewis G Rosenthal, CNA, CLP, CLE, CWTS
Rosenthal & Rosenthal, LLC www.2rosenthals.com
Need a managed Wi-Fi hotspot? www.hautspot.com
visit my IT blog www.2rosenthals.net/wordpress
-------------------------------------------------------------

Dave Saville

unread,
Oct 1, 2012, 6:41:21 AM10/1/12
to
On Sun, 30 Sep 2012 19:56:42 UTC, Lewis Rosenthal
<lgros...@2-de-sp-am-2rosenthals.com> wrote:

> Hey, Dave!
>
> On 09/30/12 02:34 pm, Dave Saville thus wrote :
> > Takes tens of seconds with the hard drive going like the clappers. WPS
> > totally un responsive until it ends.
> >
> > What the heck is it doing?
>
> Sounds like one or more of the sqlite databases needs some maintenance.
>
> We had a good discussion about this in this ng back in June. Check that
> thread for some tips (subject: SQLite vacuuming).
>
> Also, Steve has some tips on this on his Warpzilla page (look for
> SQLite): http://www.os2bbs.com/os2news/warpzilla.html

Well nothing compressed very much except places which went down from
10 meg to 1 and a bit. Have turned off all history now. Never see much
use for it.

--
Regards
Dave Saville

Dave Saville

unread,
Oct 1, 2012, 6:47:33 AM10/1/12
to
On Mon, 1 Oct 2012 10:41:21 UTC, "Dave Saville" <da...@invalid.invalid>
wrote:

> Well nothing compressed very much except places which went down from
> 10 meg to 1 and a bit. Have turned off all history now. Never see much
> use for it.
>

And as soon as I start FF it goes back up to 10 meg :-( I have all
history off - not private browsing just "custom" and nothing checked.

--
Regards
Dave Saville

Steve Wendt

unread,
Oct 2, 2012, 12:30:15 AM10/2/12
to
On 10/01/12 03:47 am, Dave Saville wrote:

>> Well nothing compressed very much except places which went down from
>> 10 meg to 1 and a bit. Have turned off all history now.
>
> And as soon as I start FF it goes back up to 10 meg :-( I have all
> history off - not private browsing just "custom" and nothing checked.

At some point, they probably set that as the minimum size, to prevent
the disk activity incurred by having to grow the file all the time.

There was a time when manually "vacuuming" the SQLite files made an
impact on performance, but I'm not sure it's still true. They did a ton
of optimization work.

Lewis Rosenthal

unread,
Oct 3, 2012, 12:19:46 AM10/3/12
to
On 10/02/12 12:30 am, Steve Wendt thus wrote :
Indeed. I do see some improvement, however, when I've been experiencing
particularly sluggish behavior, which is what got me thinking about
Dave's problem.

Dave, 10MB is about right, but deflating it and letting it rebuild was
probably a good thing, anyway.

Going back to your original issue, and thinking through it as though I
were called in to fix it, see if you get the same behavior in safe mode.
Don't rule out profile corruption (beyond the database stuff we've
discussed already; test a fresh profile entirely).

GL

Dave Saville

unread,
Oct 3, 2012, 12:11:22 PM10/3/12
to
On Wed, 3 Oct 2012 04:19:46 UTC, Lewis Rosenthal
<lgros...@2-de-sp-am-2rosenthals.com> wrote:

> On 10/02/12 12:30 am, Steve Wendt thus wrote :
> > On 10/01/12 03:47 am, Dave Saville wrote:
> >
> >>> Well nothing compressed very much except places which went down from
> >>> 10 meg to 1 and a bit. Have turned off all history now.
> >>
> >> And as soon as I start FF it goes back up to 10 meg :-( I have all
> >> history off - not private browsing just "custom" and nothing checked.
> >
> > At some point, they probably set that as the minimum size, to prevent
> > the disk activity incurred by having to grow the file all the time.
> >
> > There was a time when manually "vacuuming" the SQLite files made an
> > impact on performance, but I'm not sure it's still true. They did a ton
> > of optimization work.
>
> Indeed. I do see some improvement, however, when I've been experiencing
> particularly sluggish behavior, which is what got me thinking about
> Dave's problem.
>
> Dave, 10MB is about right, but deflating it and letting it rebuild was
> probably a good thing, anyway.
>

Before the compressing I had tried just deleting the thing but it
bounces back at 10MB. How the heck can a database with nothing in it
take up 10MB? I have all histories turned off.

> Going back to your original issue, and thinking through it as though I
> were called in to fix it, see if you get the same behavior in safe mode.
> Don't rule out profile corruption (beyond the database stuff we've
> discussed already; test a fresh profile entirely).

After everything was compressed it was fine for a couple of shutdowns
- now it's back to glue mode again.

--
Regards
Dave Saville

Felix Miata

unread,
Oct 3, 2012, 3:29:13 PM10/3/12
to
On 2012-10-03 11:11 (GMT-0500) Dave Saville composed:

> Before the compressing I had tried just deleting the thing but it
> bounces back at 10MB. How the heck can a database with nothing in it
> take up 10MB?

I don't know, and I didn't look any of this up. Call it educated speculation
of you wish, only because you kept asking.

I think it works like a swap file, except you don't get to enforce it's
initial or minimum size via CONFIG.SYS. It's there whether or not it's ever
actually needed. For efficiency reasons FF probably wants to be contiguous
and write specific portions of the file instead of the whole file as changes
get made, like any competent paging system, and like database files. On
OS2/eCS maybe it isn't afforded an opportunity via OS or disk driver or
missing or broken OS-specific code to write less than the whole file. It may
have an about:config entry to override the default size, while 10M may be the
minimum it'll obey.
--
"The wise are known for their understanding, and pleasant
words are persuasive." Proverbs 16:21 (New Living Translation)

Team OS/2 ** Reg. Linux User #211409 ** a11y rocks!

Felix Miata *** http://fm.no-ip.com/

Steve Wendt

unread,
Oct 4, 2012, 12:11:37 AM10/4/12
to
On 10/01/12 09:30 pm, Steve Wendt wrote:

>> And as soon as I start FF it goes back up to 10 meg :-(
>
> At some point, they probably set that as the minimum size, to prevent
> the disk activity incurred by having to grow the file all the time.

http://support.mozilla.org/en-US/questions/786526

Trawling Bugzilla for follow-up bugs being filed, as was suggested, is
an exercise left to those more interested:
https://bugzilla.mozilla.org/show_bug.cgi?id=581606#c37

Lewis Rosenthal

unread,
Oct 4, 2012, 1:23:06 AM10/4/12
to
On 10/03/12 03:29 pm, Felix Miata thus wrote :
> On 2012-10-03 11:11 (GMT-0500) Dave Saville composed:
>
>> Before the compressing I had tried just deleting the thing but it
>> bounces back at 10MB. How the heck can a database with nothing in it
>> take up 10MB?
>
> I don't know, and I didn't look any of this up. Call it educated
> speculation of you wish, only because you kept asking.
>
> I think it works like a swap file, except you don't get to enforce it's
> initial or minimum size via CONFIG.SYS. It's there whether or not it's
> ever actually needed. For efficiency reasons FF probably wants to be
> contiguous and write specific portions of the file instead of the whole
> file as changes get made, like any competent paging system, and like
> database files. On OS2/eCS maybe it isn't afforded an opportunity via OS
> or disk driver or missing or broken OS-specific code to write less than
> the whole file. It may have an about:config entry to override the
> default size, while 10M may be the minimum it'll obey.

Steve's links were quite interesting. It's good to know that we're not
the only ones who think that 10MB for an essentially empty file is a bit
excessive. I think one of the bugs I hit on while scanning included a
proposal for just what you mention, Felix: the ability to trim the
default/starting size of places.sqlite via a pref. Unfortunately, I
don't see that implemented as yet.

Dave, you might want to check your prefs for:

sqlite_opt.reindex_on_exit

and

sqlite_opt.auto_reindex_per_browser_close

Perhaps your shutdown issue is related to FF trying to reindex the file
and having trouble (though if you've already deleted it and had it
rebuild from scratch, I would be hard pressed to speculate as to what
could cause such apparently consistent behavior other than some failing
hardware).

Just another thought to throw into the mix.

Cheers/2

Dave Saville

unread,
Oct 4, 2012, 6:41:20 AM10/4/12
to
On Thu, 4 Oct 2012 05:23:06 UTC, Lewis Rosenthal
<lgros...@2-de-sp-am-2rosenthals.com> wrote:

<snip>
> Dave, you might want to check your prefs for:
>
> sqlite_opt.reindex_on_exit
>
> and
>
> sqlite_opt.auto_reindex_per_browser_close

All I have is

storage.vacuum.last.places.sqlite;1348220588

and I don't know where that came from as it says "user set".

Just found this
http://techlogon.com/2012/04/01/compact-firefox-databases-to-improve-p
erformance-or-not/
--
Regards
Dave Saville

Dave Saville

unread,
Oct 4, 2012, 6:57:28 AM10/4/12
to
On Thu, 4 Oct 2012 05:23:06 UTC, Lewis Rosenthal
<lgros...@2-de-sp-am-2rosenthals.com> wrote:

<snip>

> Dave, you might want to check your prefs for:
>
> sqlite_opt.reindex_on_exit
>
> and
>
> sqlite_opt.auto_reindex_per_browser_close
>

What do they do? Google can't find either.

--
Regards
Dave Saville

Paul Ratcliffe

unread,
Oct 4, 2012, 7:35:38 AM10/4/12
to
On Sun, 30 Sep 2012 13:34:47 -0500, Dave Saville <da...@invalid.invalid> wrote:

> Takes tens of seconds with the hard drive going like the clappers. WPS
> totally un responsive until it ends.
>
> What the heck is it doing?

The real question is why is it doing it on a user interface thread
blocking the message queue? It should be doing it on a worker thread
and then you wouldn't care.

I'm afraid the FF developers have lost the plot. They don't even seem
to know what goes on in their own code, or how or why, so how are the
users supposed to second guess it?

Dave Yeo

unread,
Oct 4, 2012, 11:23:45 AM10/4/12
to
Lewis Rosenthal wrote:
> On 10/03/12 03:29 pm, Felix Miata thus wrote :
>> On 2012-10-03 11:11 (GMT-0500) Dave Saville composed:
>>
>>> Before the compressing I had tried just deleting the thing but it
>>> bounces back at 10MB. How the heck can a database with nothing in it
>>> take up 10MB?
>>
>> I don't know, and I didn't look any of this up. Call it educated
>> speculation of you wish, only because you kept asking.
>>
>> I think it works like a swap file, except you don't get to enforce it's
>> initial or minimum size via CONFIG.SYS. It's there whether or not it's
>> ever actually needed. For efficiency reasons FF probably wants to be
>> contiguous and write specific portions of the file instead of the whole
>> file as changes get made, like any competent paging system, and like
>> database files. On OS2/eCS maybe it isn't afforded an opportunity via OS
>> or disk driver or missing or broken OS-specific code to write less than
>> the whole file. It may have an about:config entry to override the
>> default size, while 10M may be the minimum it'll obey.
>
> Steve's links were quite interesting. It's good to know that we're not
> the only ones who think that 10MB for an essentially empty file is a bit
> excessive. I think one of the bugs I hit on while scanning included a
> proposal for just what you mention, Felix: the ability to trim the
> default/starting size of places.sqlite via a pref. Unfortunately, I
> don't see that implemented as yet.

It does make sense to avoid file system fragmentation which I understand
is still an issue under Windows.
Doesn't seem to be a pref but the default goes away if the amount of
free space is less then 500 MBS

>
> Dave, you might want to check your prefs for:
>
> sqlite_opt.reindex_on_exit
>
> and
>
> sqlite_opt.auto_reindex_per_browser_close
>
> Perhaps your shutdown issue is related to FF trying to reindex the file
> and having trouble (though if you've already deleted it and had it
> rebuild from scratch, I would be hard pressed to speculate as to what
> could cause such apparently consistent behavior other than some failing
> hardware).
>
> Just another thought to throw into the mix.
>

I doubt it is anything to do with places.sqlite and Dave should try
safe-mode and a different profile. Also maybe look at the sessionstore
files.
Dave

Steve Wendt

unread,
Oct 4, 2012, 12:02:35 PM10/4/12
to
On 10/4/2012 3:41 AM, Dave Saville wrote:

> Just found this
> http://techlogon.com/2012/04/01/compact-firefox-databases-to-improve-performance-or-not/

That appears to answer all of your questions, other than the original
one, which doesn't seem to be a problem others are seeing. Hence the
suggestion to try a new profile.

Lewis Rosenthal

unread,
Oct 5, 2012, 9:46:57 PM10/5/12
to
Sorry for the late follow-up...

On 10/04/12 06:57 am, Dave Saville thus wrote :
Hmmm... Must have been installed as part of
https://addons.mozilla.org/en-us/firefox/addon/sqlite-optimizer/ . I
thought they were standard prefs; apologies (I had to look again).

Essentially, they just add some auto-compacting to the db engine, but
I'm really leaning toward some profile corruption, as the more I think
about this, the harder it is for me to come up with another system I've
*ever* seen on *any* platform which has become so
fragmented/bloated/etc. that it has behaved as your seems to be.
0 new messages