Occasional 'Object graph cannot be null' error

[s6521d]

Hi, I'm using File Cache as the persistence store for LazyCache and very occasionally when reading a cachce key I'm see an 'Object graph cannot be null' error and the file for the cache key on disk is 0 bytes in length.

You'll see from the code snippet in the attached that I've used lock(){} to try and rule out and threading issues.
Object graph cannot be null - 2.pdf

Any help much appreciated.

[acarteas]

Hmm. That's weird. Do you have any logs that give a better clue as to under what circumstances this happens? Regardless, I'll take a look at the code and see if I can somehow control for / trap this error and maybe return an empty string instead.

…

On Wed, Jul 4, 2018 at 8:43 AM, s6521d ***@***.***> wrote: Hi, I'm using File Cache as the persistence store for LazyCache and very occasionally when reading a cachce key I'm see an 'Object graph cannot be null' error and the file for the cache key on disk is 0 bytes in length. You'll see from the code snippet in the attached that I've used lock(){} to try and rule out and threading issues. Any advise would be most appreciated Object graph cannot be null.pdf <https://github.com/acarteas/FileCache/files/2163753/Object.graph.cannot.be.null.pdf> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#18>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AP2PoyNvyBbG6nsrMFupasA4mNPA7jQfks5uDOK0gaJpZM4VCyBg> .

[s6521d]

Sorry no logs, what would you suggest?

[acarteas]

I'm not sure. Can you confirm that every call to Get is requesting a valid cache key?

…

On Fri, Jul 6, 2018 at 5:26 AM, s6521d ***@***.***> wrote: Sorry no logs, what would you suggest? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#18 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AP2Poy_oAILUKsfYnn7EPHW5LNg8S9Xmks5uD1dggaJpZM4VCyBg> .

[s6521d]

Sorry for the delay, yes all calls are using a valid cache key.
Is there anything around thread safety that we need to be aware of when using FileCache?

[s6521d]

I've just written a console application that uses two threads to read the same key from a static instance LazyCache backed by file cache and I very quickly see the Object graph cannot be null' error.

Update: added a 25ms Thread.Sleep after the read and the this error has gone away

Update: increasing threads to 4 and the error is back

Happy to share my code with you.

[acarteas]

That'd be great. Also, try the latest version of the source on Github (not posted to nuget yet). It includes a PR that fixes a bug related to file accessing.

[s6521d]

Hi,  Sorry for the delay in replying to you. I'm having issues building the current source, is there any chance you zip it up built so I can add the dll file to my test solution? Kind regards, Steve Dawkins On Monday, 17 September 2018, 19:00, Adam Carter <notifications@github.com> wrote: That'd be great. Also, try the latest version of the source on Github (not posted to nuget yet). It includes a PR that fixes a bug related to file accessing.— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[s6521d]

Hi I've got the source to build and now my test program fails with 'The process cannot access the file' error rather than any 'Object graph cannot be null' errors; zip of test program is attached. The error is not on every read attempt but there's obviously still some file contention when the cache is under a heavy read load.
DfE-403.zip

The number after EXCEPTION is a read count and the number in [ ] is the Thread.CurrentThread.ManagedThreadId

[s6521d]

Sorry closed in error

[acarteas]

Okay, I'll take a look soon and get back to you.

…

On Mon, Sep 24, 2018 at 8:24 AM s6521d ***@***.***> wrote: Sorry closed in error — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#18 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AP2Po_vQuvIYC9p-fqchPZudkYFeFvuKks5uePkkgaJpZM4VCyBg> .

[acarteas]

I think that found the culprit on line 969. Since FC is doing a read a few lines before, I'm not sure if it's necessary to force updating last access time. I'm going to do some regression testing to make sure removing that doesn't break anything. If all looks good, I'll push the changes soon.