search

michaelogrant / macfusion

Automatically exported from code.google.com/p/macfusion
Other
1 stars 1 forks source link

Unmount on Sleep, Remount on Wake - a MacFusion *default* please (to minimise risk of data loss) #25

Closed GoogleCodeExporter closed 8 years ago

GoogleCodeExporter commented 8 years ago 
With MacFusion Preferences for 
Sleep/Wake Behaviour set to 

> Don't touch mounts 

System Preferences… | Location | Automatic

Working from home, connected via Ethernet. 
UK Online broadband. NAT but in DMZ. 

Mount two volumes via ssh.

Word documents on each volume.

Word 2004 opened one document from each location. 

Word 2004 preferences are already:

* auto save disabled
* fast save disabled.

Edited both documents, saved changes successfully. 

Edited both documents, did not save changes. 

Closed lid of MacBook Pro. Allowed time for sleep routine to complete (watched 
for pulsing LED). 

Disconnected from Ethernet. 

Drove to office. 

Connected to Ethernet. 

Connected Cinema HD display to DVI port.

Lifted lid. 

IP address 139.184.224.243 obtained from DHCP server 
(observed using IP in menu bar 
<http://www.monkeybreadsoftware.de/Freeware/index.shtml>). 

Apple menu | Location | University of Sussex 
(preparing for when I disconnect from wired Ethernet; our 'Roaming' Wi-Fi 
network will kick in 
and require a PAC file URL). 

IP in menu bar reported, for a few moments: 

> Not connected!

IP in menu bar reported 139.184.224.243

Brought Word to foreground. 

Command-s to save. 

For both volumes mounted via ssh, 
MacFUSE file system daemon stopped responding. 

I continued to wait for a reasonable period (five minutes, maybe) but 
ultimately forced both to 
Eject. 

###

Word did not crash; rather, it prompted to save the files to alternative 
locations. I did not follow 
this prompt ... I *suspect* that if I did so, Word may have stopped responding 
(as it attempted to 
read Word Work Files and/or temporary files from an unavailable volume). 

###

From the routine above, I can't estimate whether the issue arose before or 
after I changed 
locations. 

Question 1
----------

Should I prefer something other than 

> Don't touch mounts 

for Sleep/Wake Behaviour? 

Question 2
----------

Is there an advanced option (without a GUI in the current beta of MacFusion) 
that might avoid 
this issue?

###

If the above is indicative of an issue with MacFUSE/sshfs-static, I'll create a 
new issue in that 
area.

If however there's an option (or advanced option) within MacFusion that should 
be a default to 
avoid this problem, then I'll suggest this as an enhancement -- with a warning 
presented to the 
user, should they prefer stray from the MacFusion default.

Thanks
Graham

Original issue reported on code.google.com by grahampe...@gmail.com on 1 May 2007 at 8:11

Attachments:

GoogleCodeExporter commented 8 years ago 
macbookpro03:~ gjp22$ cd /Volumes/
macbookpro03:/Volumes gjp22$ ls -l
total 8
lrwxr-xr-x   1 root  admin  1 May  1 02:38 MBP -> /
d---------   1 root  wheel  0 Jan  1  1970 keeler
d---------   1 root  wheel  0 Jan  1  1970 omnium-gjp22

I used Finder eject to remove from 
/Volumes
the directories that had not removed following use of the Eject button 
presented by MacFusion. 

macbookpro03:/Volumes gjp22$ ls -l
total 8
lrwxr-xr-x   1 root  admin  1 May  1 02:38 MBP -> /

Original comment by grahampe...@gmail.com on 1 May 2007 at 8:15

GoogleCodeExporter commented 8 years ago 
Used MacFusion to re-mount both volumes. 

Consider the directory listing for keeler, which hosts one of the two test 
documents that I had been editing:

macbookpro03:/Volumes gjp22$ ls -alt keeler/
total 8408
drwxrwxrwt   6 root   admin      204 May  1 09:17 ..
drwx------   1 gjp22  111       1024 May  1 07:33 Mail
-rw-r--r--   1 gjp22  111         82 May  1 07:32 ._Winnov queries copy.doc
drwx--S---   1 gjp22  111       2048 May  1 07:31 .
-rw-------   1 gjp22  111       4036 Apr 30 19:31 .sh_history
-rw-r--r--   1 gjp22  111      75326 Apr 30 10:53 2007-04-30 welcome Rosie 
Mulgrue.pdf
drwxr-xr-x   1 gjp22  111         96 Apr 28 11:43 159
-rw-------   1 gjp22  111      12292 Apr 28 11:30 .DS_Store
-rw-r--r--   1 gjp22  111         82 Apr 27 04:00 ._JORUM_osswatch_final.doc
-rw-r--r--   1 gjp22  111    3871232 Apr 27 04:00 JORUM_osswatch_final.doc
-rw-r--r--   1 gjp22  111         82 Apr 20 12:26 ._Winnov queries.doc
-rw-r--r--   1 gjp22  111      53760 Apr 20 12:26 Winnov queries copy.doc
-rw-r--r--   1 gjp22  111      53760 Apr 20 12:26 Winnov queries.doc
-rw-r--r--   1 gjp22  111        455 Apr 14 14:25 .mlbxlsttmp

[snip]

I find it curious that I had successfully used Word to save changes this 
morning to 
Winnov queries copy.doc
but only 
._Winnov queries copy.doc 
bears today's date.

Original comment by grahampe...@gmail.com on 1 May 2007 at 8:26

GoogleCodeExporter commented 8 years ago 
A comparison

The time-sorted listing for the directory (on the other server) that hosted the 
other document presents a 
slightly different picture...

macbookpro03:/Volumes/omnium-gjp22/Documents gjp22$ ls -alt | more
total 9136
drwx--x--x   1 gjp22  staff     3672 May  1 07:32 .
-rw-r--r--   1 gjp22  staff       82 May  1 07:32 ._2005-11-05 Windows XP 
post-installation errors copy.doc
-rw-r--r--   1 gjp22  staff   395264 May  1 07:32 2005-11-05 Windows XP 
post-installation errors copy.doc
-rw-------   1 gjp22  staff       82 May  1 07:31 ._Word Work File L_3
drwxr-xr-x   1 gjp22  staff     2142 Mar  6 18:57 ..
-rw-r--r--   1 gjp22  staff       82 Mar  3 12:09 ._Hello again.doc

[snip]

That seems more logical, both 
._2005-11-05 Windows XP post-installation errors copy.doc
and 
2005-11-05 Windows XP post-installation errors copy.doc
modified this morning.

Original comment by grahampe...@gmail.com on 1 May 2007 at 8:29

GoogleCodeExporter commented 8 years ago 
http://episteme.arstechnica.com/eve/forums/a/tpc/f/8300945231/m/394001754831?
r=687008984831#687008984831

> auto unmount on sleep is a wonderful, wonderful thing? 
> No more beachballs

Original comment by grahampe...@gmail.com on 11 May 2007 at 1:08

GoogleCodeExporter commented 8 years ago 
See also MacFUSE issue 
http://code.google.com/p/macfuse/issues/detail?id=173

Original comment by grahampe...@gmail.com on 11 May 2007 at 11:14

GoogleCodeExporter commented 8 years ago 
MacFusion issue 86 regarding limitations of sshfs is a 

    HIGHLIGHT

in the context of this issue 25 
and other novel/unexpected uses of the sshfs system in general 
(not the sshfs-static binary in particular).

Original comment by grahampe...@gmail.com on 12 May 2007 at 10:15

GoogleCodeExporter commented 8 years ago 
I opted to test with 

> auto unmount on sleep 

and an application that is likely to be less stressful than Microsoft Word 
2004. 

1)  Preparing for my journey from home to office

2)  Apple menu | Location | I changed location

3)  I didn't watch for the brief disconnection that is usually observed 
    in 'IP in menu bar' (at this point in this test, a disconnection would probably be irrelevant)

4)  I waited for a moment

5)  MacFusion 1.1 Beta 2 | Favourites | omnium-gjp22 
    (an ssh connection to my home directory on that server)

6)  TextWrangler | File menu | Open recent | 
    index.html - /Volumes/omnium-gjp22/Sites/2007/05/13

7)  I edited the file

8)  I did not opt to save

9)  Apple menu | Sleep

10) I waited for the LED on my MacBook Pro 17" to begin pulsing

11) I closed the lid, LED continued to pulse

12) I removed the Ethernet cable, LED continued to pulse

13) I journeyed to work

14) I connected my Cinema HD display to the DVI port

15) I connected an Ethernet cable

16) I connected the power adapter

17) at this point the laptop awoke

18) TextWranger | command-s | changes were saved to the file :)

19) I command-clicked the proxy icon in the title bar of the file

20) I released at /Volumes/omnium-gjp22/Sites/2007/

21) TextWrangler | command-w

22) I opened the lid

23) display menu icon | Detect Displays

24) TextWrangler | File menu | Open recent | 
    index.html - /Volumes/omnium-gjp22/Sites/2007/05/13

25) I edited, saved changes, success :)

After each saves I browsed to 
<http://omnium.freeman-centre.ac.uk/~gjp22/2007/05/13/> and reaffirmed that 
saved changes to 
index.html were apparent. They were. 

SUCCESS :)

My next step should be to test using something more stressful. Microsoft Word 
comes to mind.

Original comment by grahampe...@gmail.com on 14 May 2007 at 8:44

GoogleCodeExporter commented 8 years ago 
Priority-Critical 
as my current sense is that if 

  Unmount on Sleep, Remount on Wake

is NOT a default

THEN there may be a greater risk of users finding themselves unable to save 
changes -- to documents that 
they were editing before a computer slept -- after the computer has woken. 

We already have
Label:Component-Docs 
— when we're more certain about this (I have some more testing to do) we 
might provide documentation/Help 
warning against (and better explaining the possible implications of) certain 
Sleep/Wake preferences.

Original comment by grahampe...@gmail.com on 14 May 2007 at 8:48

GoogleCodeExporter commented 8 years ago 
We (I) should also check to see what might happen as a result of the brief 
disconnection that sometimes occurs 
when a location is changed (in software). For example: will a brief 
disconnection such as that have an effect on 
unsaved changes to Microsoft Word documents? (Could be a separate issue in due 
course.)

Original comment by grahampe...@gmail.com on 14 May 2007 at 8:51

GoogleCodeExporter commented 8 years ago 
Since I have preferred for MacFusion to 

  Unmount on Sleep, Remount on Wake

I can no longer reproduce the issue of Word (or TextWrangler) failing to save 
changes to files that were edited 
(but unsaved) prior to sleep. 

As a critical suggestion (to avoid loss of data by users): 

  Unmount on Sleep, Remount on Wake

should be the default, with explanatory Help/documentation. 

Unless, that is, there's a good reason to default to some other sleep 
behaviour. 

Thanks

Original comment by grahampe...@gmail.com on 15 May 2007 at 9:38

GoogleCodeExporter commented 8 years ago 
For reference only: here are the notes that I entered into the Word document as 
I was editing, saving etc. 
during the test: 

1.  tt connecting via SFTP (sshfs) to omnium, MacFusion 1.1 Beta 2
2.  editing this document with Word 2004 on Mac OS X 10.4.9
3.  Save — successful
4.  OK so far
5.  edit some more
6.  sleep
7.  Growl notified me that

    *   MacFusion failed to Unmount the volume

8.  displays went blank
9.  pulsing LED to confirmed sleep
10. disconnected from Ethernet
11. reconnected to Ethernet
12. wake
13. Save — successful
14. edit some more
15. intention:

    a.  sleep, disconnect, reconnect, wake
    b.  change location
    c.  observe the brief disconnection (IP in menu bar application) 
        that is typical for this particular change of location (from Automatic to Freeman Centre)
    d.  attempt save as soon as IP in menu bar indicates that 
        connection is re-established…
    e.  SUCCESS!

Original comment by grahampe...@gmail.com on 15 May 2007 at 9:45

GoogleCodeExporter commented 8 years ago 
For the benefit of any reader who is alarmed by 

>   MacFusion failed to Unmount the volume

in my previous comment, that is separated into issue 109 (apparently, not a 
defect).

Original comment by grahampe...@gmail.com on 15 May 2007 at 10:04

GoogleCodeExporter commented 8 years ago 
This morning I performed similar tests and had to wait maybe ten minutes, 
following a wake and a (soft) 
change of network location, for unresponsiveness to end. Details to follow and 
in the meantime, for the 
record: 

<http://pastie.textmate.org/62024> (sshfs-static open files and ports during 
*unresponsive* time)
<http://pastie.textmate.org/62025> (sshfs-static open files and ports *after* 
responsiveness returned)
<http://pastie.textmate.org/62023> (sshfs-static sample *after* responsiveness 
returned)
<http://pastie.textmate.org/62026> (extract from console.log
<http://pastie.textmate.org/62027> (extract from system.log)

Attached: a debug log that may assist with any one of, or combination of, 
MacFusion 
issue 25
issue 109
issue 116

The Word document with which I was testing was ultimately saved successfully, 
and I Quit from Word with no 
apparent problem. 

The lengthy period of unresponsiveness was undesirable but as it was ultimately 
possible to save changes to 
my Word document, and to Quit from Word thereafter, at this point I still veer 
towards 

 Unmount on Sleep, Remount on Wake

as a default.

Original comment by grahampe...@gmail.com on 16 May 2007 at 11:11

Attachments:

GoogleCodeExporter commented 8 years ago 
See also 
MacFusion issue 118

Original comment by grahampe...@gmail.com on 16 May 2007 at 12:07

GoogleCodeExporter commented 8 years ago 
Again, this morning, I performed tests. On both occasions saves were 
successful. 

This morning's successful tests with Word were followed by: 

* a very long wait after I clicked the eject icon against the volume in Finder 
sidebar (Continue to Wait etc. in 
the dialogue from MacFUSE level)

* Activity Monitor and IP in menu bar (application) unresponsive during that 
wait (issue 118)

* a requirement to Force Eject, and continued running of the sshfs-static 
process (issue 119). 

The most remarkable thing was the illogically short length of time that it took 
for Word to open a large 
document over a relatively slow connection. For this, I'll raise a separate 
issue. 

(Whilst focusing on Word, I don't forget the utility and limitations of SSHFS 
as a system.)

Original comment by grahampe...@gmail.com on 16 May 2007 at 5:45

GoogleCodeExporter commented 8 years ago 
Re my comment 13 above, here are three snapshots that I took during that wait 
(for responsiveness, after 
waking, at the MacFUSE level) before a save was possible. 

I remind myself that at points during the wait (in my judgement, at that time, 
*not* too soon after the wait 
had begun) I experimented by:

  # enabling AirPort (without letting it become the primary connection, to the Internet)

  # then removing the Ethernet cable (allowing AirPort to become my connection to the Internet). 

Key points (?) 

  * Ethernet cable was connected, and AirPort disabled, 
    prior to waking the computer

  * I waited for a reasonable length of time with a 
    fast (100 Meg) link to the same switch (Z-PoP) to which 
    the server (Mac OS X Server 10.3.9) is connected

  * some time (I shan't estimate how long) after 
    enabling AirPort then removing the Ethernet cable 
    (allowing AirPort to make my connection to the Internet) 
    Word became responsive and allowed me to save changes 
    without difficulty.

Original comment by grahampe...@gmail.com on 16 May 2007 at 6:11

Attachments:

GoogleCodeExporter commented 8 years ago 
Sleep wake issues to be address before 1.1.

Original comment by mgorb...@gmail.com on 17 May 2007 at 3:41

GoogleCodeExporter commented 8 years ago 
I'll test again very shortly. 

In the meantime I'm bookmarking  
<http://www.macosxhints.com/article.php?story=20070517224112286> 
which discusses the 
kicker
daemon.

Original comment by grahampe...@gmail.com on 22 May 2007 at 1:06

GoogleCodeExporter commented 8 years ago 
This is now the ONLY blocker bug on 1.1.
Hopefully Graham and Michael can talk tomorrow to resolve this, and 1.1 is go.

Original comment by mgorb...@gmail.com on 22 May 2007 at 4:00

GoogleCodeExporter commented 8 years ago 
I'll test once more in a few moments, 

> Unmount on Sleep, Remount on Wake

and report back.

Original comment by grahampe...@gmail.com on 22 May 2007 at 9:23

GoogleCodeExporter commented 8 years ago 
As I suspected, it does seem that working with Microsoft Office documents in 
general (I shan't limit my 
cautions to Word 2004) is a lottery. 

Ultimately on this occasion 

> Word failed reading from this file...
> [Cancel] [[Retry]]

And when I eventually Cancelled (gave up retrying), long after the connection 
to the ssh connection to the 
remote system had been restored, Word crashed. 

Word crash report is at <http://pastie.textmate.org/63499> and I didn't bother 
sending it to Apple. 

If we begin to fine tune sshfs, MacFUSE or MacFusion for a moving and (IMO) 
poorly documented target such 
as Word, then we place ourselves on the edge of a precipice. 

My sense is that we should acknowledge -- and present clearly to the end user 
-- the limitations of  
_systems_ such as sshfs in a FUSE environment (and explain, without 
finger-pointing, that it's not a problem 
with MacFUSE, sshfs-static or MacFusion).

Let's discuss :)

Original comment by grahampe...@gmail.com on 22 May 2007 at 10:30

GoogleCodeExporter commented 8 years ago 
> a moving and (IMO) poorly documented target such as Word

Brace yourselves. In an MSDN blog 
<http://blogs.msdn.com/rick_schaut/archive/2004/05/19/135315.aspx> 
someone explains why it can be so difficult to identify and fix bugs.

Original comment by grahampe...@gmail.com on 22 May 2007 at 10:41

GoogleCodeExporter commented 8 years ago 
Per conversation with Graham, tagging as WONTFIX right now.
To be addressed later.

Original comment by mgorb...@gmail.com on 22 May 2007 at 2:40

GoogleCodeExporter commented 8 years ago 
Avoid a long system lockup related to sleeping shares
<http://www.macosxhints.com/article.php?story=20070522053721530>
-- I don't know whether anything there will be relevant in the 
MacFUSE/MacFusion contexts but at least, there 
are people interested in the subject generally.

Original comment by grahampe...@gmail.com on 23 May 2007 at 9:30

GoogleCodeExporter commented 8 years ago 
See also MacFusion issue 145.

Original comment by grahampe...@gmail.com on 23 May 2007 at 1:23

GoogleCodeExporter commented 8 years ago 
At 
<http://episteme.arstechnica.com/eve/forums/a/tpc/f/8300945231/m/341003745831> 
<http://episteme.arstechnica.com/eve/forums/a/tpc/f/8300945231/m/800002745831/p/
2> and elsewhere 
there are differing interpretations of what autofs will bring to Mac OS X 10.5. 

At <http://www.apple.com/macosx/leopard/technology/unix.html>:

> The Finder and other applications needn’t wait for one mount to
> complete before requesting another

-- whilst some people may understand this to mean 'no more Finder beach balls' 
in the broadest sense, I 
focus on the word 'mount'. 

When we sleep our computers and/or change networks — necessarily disrupting 
SSH and other connections — 
there's much more going on than a simple mount. 

I wonder: should we _expect_ file systems such as SSHFS, which depend upon 
networking, to behave 
gracefully when the network is disrupted?

(If behaviour is graceful, that's a bonus, but should we _expect_ it?)

Bearing in mind: 

* SSHFS, CurlFtpFS etc. are described as 'utility' file systems

* historically, protocols such as AFP have greater consideration for network 
disruption etc..

<http://groups.google.com/group/MacFusion-devel/browse_frm/thread/
5b412e06839f7609/479222d09de746a8> invites discussion.

Original comment by grahampe...@gmail.com on 17 Jun 2007 at 8:14

GoogleCodeExporter commented 8 years ago 
At <http://ircbrowse.com/channel/macfusion/20070618#hour23> positive comments 
re: wake from sleep

Original comment by grahampe...@gmail.com on 18 Jun 2007 at 10:59