• 0

FileStream - The Process Cannot Access The File


Question

Hello, 

 

I have a file copying program and I have a custom class that utilizes a System.IO.FileSteam to copy a file to a destination computer.  For the most part it works quite well.  However I use it to copy to many remote locations.  Some of them have subpar network connections.  Every so often I'll get a server that has a unreliable network connection and it will momentarily drop off the network for a few seconds.  When this happens my FileStream breaks with the exception "The specified network name is no longer available.".  This is expected.

However the problem is after the program has waited 60 seconds it will retry the transfer.  The problem now is it can't even start because I get the error "The process cannot access the file '\\MyServer\c$\MyFile.zip' because it is being used by another process".  I'll continue to get this error indefinitely until I close the program and restart the transfer, then it is able to transfer once again.

It would appear that my source server still has a file handle open when this happens that's preventing the transfer again and closing the program releases it.  I would like to programically release the handle so I can restart the transfer automatically.

 

Here's what I've tried so far:

When the file transfer is started, I grab the FileSteam.SafeFileHandle to a variable.  When I get the above error, I've tried accessing SafeFileHandle.SetHandleAsInvalid(), SafeFileHandle.DangerousRelease() and SafeFileHandle.Close(), but I'll still get the file in use error when I retry.

12 answers to this question

Recommended Posts

  • 0

To replicate the problem, I imported my DLL into powershell.  I waited for the problem to occur.  I then grabbed the SafeFileHandle.  I first tried SetFileHandleAsInvalid() then retried.  Then I tried DangerousRelease().  Finally I tried Close(), but at that point I got "Exception calling "Close" with "0" argument(s): "Safe handle has been closed".

 

Also if I just call my handle variable it shows:   IsInvalid:  False, IsClosed:  True

  • 0
  On 13/03/2014 at 21:48, snaphat (Myles Landwehr) said:

Is the filestream locked? Closing a file that has outstanding locks is undefined according to the documentation.

 

Some how I never thought of that.  I always assumed that when I received an exception on a FileStream.Write(...) that the filestream was hosed then.  I adjusted my code as follows:

                try
                { m_Dest.Write(m_Buffer, 0, m_CurrentBlockSize); }
                catch (IOException ex)
                {
                    //  added the section below:
                    if (m_Dest.CanWrite)
                    {
                        try
                        { m_Dest.Close(); }
                        catch { }
                    }
                    // end new section
                    HadExceptionWhileCopying(ex, true);
                    return;
                }

So far it worked in my test environment.  Will see if it works in production tomorrow.

Thanks Myles!

  • 0

^ yeah the filestream may still be valid and be locked. For that matter, it could technically be still not working (in terms of communication), but it could still be marked as locked. I was thinking of calling unlock() when I posted my response. I'm not sure why the above code would work given that canWrite should just give you the state of whether the stream is closed or not and if you call close() on an already closed stream it should work fine.

  • 0
  On 15/03/2014 at 00:06, stumper66 said:

My above solution worked.

 

I dug through my existing code and it was supposed to call a close on the filestream  in the event of an error but due to a flaw in my logic it never got called.

I see, that definitely fits the bill for the error you were getting!

  • 0
  On 15/03/2014 at 00:28, notchinese said:

Is it feasible given the code design to just wrap the filestream in a using() {} block?

 

"using" is an assurance that Dispose() will be called to free resources. It doesn't necessarily call Close(). try...finally would do that.

  • 0
  On 21/03/2014 at 12:54, Eric said:

"using" is an assurance that Dispose() will be called to free resources. It doesn't necessarily call Close(). try...finally would do that.

Yes it does. http://stackoverflow.com/questions/911408/does-stream-dispose-always-call-stream-close-and-stream-flush

 

In general, if an object implements Dispose() then you can expect that calling that is enough to clean it up properly, whatever the state it was in.

 

Also, using actually expands to try-finally so it's strictly equivalent, if all you're doing in the finally clause is to call Dispose().

  • 0
This topic is now closed to further replies.
  • Posts

    • Because Win7 was beautiful, much faster and more functional. Win10 (glossing over 8 as many do) was slightly faster in some cases, more functional in some cases, but some people such as myself hated how it looked and decided it wasn't worth the upgrade. Some people liked (or were ok with) the look, and thus it is a good upgrade. Win11 is like 10, but is less functional. It literally has nothing going for it, and I use it every day at work so I'm quite familiar with it.
    • I switched my mom from Chrome to Firefox and she had a serious meltdown. She even managed to figure out how to reinstall Chrome, which really surprised me. What finally got her to switch was Chrome no longer being supported on Win7 and me putting a Chrome skin on FF, and setting it up identically.
    • Feels very much like most other gnome based Linux distros. There is minimal amounts that are influenced by Windows 11, maybe just enough to make people who are switching comfortable enough with the idea. As far as I can tell its mainly just turning the 'taskbar' panel as a 100% sized static panel, rather than the default dynamic sized. Turning it from the Mac OS Dock into the Windows taskbar. The Arc Menu - that I assume you're taking not with from the screenshots, is indeed the Windows 11 style one, but it has lots of other options too, from the more traditional gnome, Windows 7 etc. Still free to install what ever Window Manager you want once you're comfortable enough with Linux though.
    • Wow, and here I'm still happily using 1080p...
    • Added an extra filter to Fail2Ban.  I thought about just adding this to my existing aibots filter, but for the time being I'm keeping it separate because it's "possible" real humans may trigger this one so as long as it doesn't start filling my inbox I'd like to get notified about these so I can adjust it as necessary in the future. I'm still holding close to 10k unique IP addresses at any given time that have been banned via the "aibots" filter that looks for certain user agent strings of known AI scrapers.  However, I've been getting an increasing amount of traffic trying to scrape the site with sanitized user agent strings that just look like normal web browsers, however... Because I enabled authentication I can now see that they're racking up lots of 401 (unauthorized) responses in the Apache "access.log" file, but they're not triggering anything in the Apache "error.log" file, which is where failed attempts to log in would appear.  Basically, if an actual human tried to log in with an invalid username and password they don't immediately go into "access.log" as a 401, they go into "error.log" with a status message such as "user FOO not found".  The only way to trigger a 401 simply by visiting the site, as far as I'm aware, is to hit "Cancel" on the login prompt, or otherwise try to access files directly without properly authenticating. So, given the fact I'm getting a few thousand 401 errors a day from sanitized user agent strings that don't show up in "error.log", which means no attempt at logging in properly, I added another jail/filter set to Fail2Ban to immediately ban anybody who triggers a 401.  This feels a bit nuclear so I may need to adjust it in the future, but as far as I'm aware so far no real humans are being inconvenienced so all I'm doing is wasting the time of some AI scraper bots. Example log entry 61.170.149.70 - - [25/Jun/2025:20:01:04 -0400] "GET /content/mdwiki_en_all_maxi_2024-06/A/Neuroregeneration HTTP/1.1" 401 3287 "https://kiwix.marcusadams.me/content/mdwiki_en_all_maxi_2024-06/A/Neuroregeneration" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/114.0.0.0 Safari/537.36 Edg/114.0.1823.43" Contents of /etc/fail2ban/filter.d/apache-401repeat.conf #Fail2Ban filter for bots and scrapers that try to access #files directly without entering credentials for apache2-auth #and therefore trigger lots of 401 errors without triggering #the apache-auth jail. # #Marcus Dean Adams [Definition] failregex = ^<HOST> .+\" 401 \d+ .*$ Contents of /etc/fail2ban/jail.d/apache-401repeat.local [apache-401repeat] enabled = true ignoreip = 10.1.1.1 port = 80,443 filter = apache-401repeat maxretry = 1 bantime = 672h findtime = 10m logpath = /var/log/apache2/access.log Oh, and all this traffic is AFTER I explicitly banned Alibaba's IP ranges that were absolutely blowing me up day and night. Observation; two of the IP addresses that have triggered this jail in the 30 or so minutes since I turned it on were owned by Microsoft.  Wonder if they're doing their own AI scraping/probing, or if that's just an Azure VM owned by somebody else.
  • Recent Achievements

    • Rising Star
      Phillip0web went up a rank
      Rising Star
    • One Month Later
      Epaminombas earned a badge
      One Month Later
    • One Year In
      Bert Fershner earned a badge
      One Year In
    • Reacting Well
      ChrisOdinUK earned a badge
      Reacting Well
    • One Year In
      Steviant earned a badge
      One Year In
  • Popular Contributors

    1. 1
      +primortal
      552
    2. 2
      ATLien_0
      208
    3. 3
      +FloatingFatMan
      175
    4. 4
      Michael Scrip
      152
    5. 5
      Som
      139
  • Tell a friend

    Love Neowin? Tell a friend!