Intermittent 'OSError: [Errno 2] No such file or directory' over network share

Question

I'm running several scripts which are writing and reading from the same CSV file over a network share.

I'm getting intermittent OSError: [Errno 2] No such file or directory: 'filename.csv' (every minute or two, with writing/reading every 10-30 seconds), but most of the time the file writing/reading works fine.

Is there any way to track which script has the file open when, and get some more details about the error? I.e. is file in use? Is it losing connection to the file server?

---

Using Process monitor:

Ordinary operation (no error messages):

2:33:52.4823608 PM  python.exe  2668    IRP_MJ_QUERY_VOLUME_INFORMATION filename.csv    SUCCESS Type: QueryDeviceInformationVolume, DeviceType: Disk, Characteristics: Remote
2:33:52.4823848 PM  python.exe  2668    IRP_MJ_QUERY_VOLUME_INFORMATION filename.csv    SUCCESS Type: QueryDeviceInformationVolume, DeviceType: Disk, Characteristics: Remote
2:33:52.4824062 PM  python.exe  2668    IRP_MJ_QUERY_VOLUME_INFORMATION filename.csv    BUFFER OVERFLOW Type: QueryInformationVolume, VolumeCreationTime: 1/1/1601 8:00:00 AM, VolumeSerialNumber: D020-FD78, SupportsObjects: False, VolumeLabel: DAT`
2:33:52.4827061 PM  python.exe  2668    IRP_MJ_QUERY_INFORMATION        filename.csv    BUFFER OVERFLOW Type: QueryAllInformationFile, CreationTime: 4/24/2012 2:36:07 PM, LastAccessTime: 4/24/2012 2:36:07 PM, LastWriteTime: 4/24/2012 2:36:07 PM, ChangeTime: 4/24/2012 2:36:07 PM, FileAttributes: A, AllocationSize: 1,048,576, EndOfFile: 118,342, NumberOfLinks: 1, DeletePending: False, Directory: False, IndexNumber: 0x45060de, EaSize: 0, Access: None, Position: 0, Mode: , AlignmentRequirement: Byte
2:33:52.4837725 PM  python.exe  2668    FASTIO_QUERY_INFORMATION        filename.csv    FAST IO DISALLOWED  Type: QueryStandardInformationFile
2:33:52.4837820 PM  python.exe  2668    IRP_MJ_QUERY_INFORMATION        filename.csv    SUCCESS Type: QueryStandardInformationFile, AllocationSize: 1,048,576, EndOfFile: 118,342, NumberOfLinks: 1, DeletePending: False, Directory: False
2:33:52.4841324 PM  python.exe  2668    FASTIO_QUERY_INFORMATION        filename.csv    FAST IO DISALLOWED  Type: QueryStandardInformationFile
2:33:52.4841458 PM  python.exe  2668    IRP_MJ_QUERY_INFORMATION        filename.csv    SUCCESS Type: QueryStandardInformationFile, AllocationSize: 1,048,576, EndOfFile: 118,342, NumberOfLinks: 1, DeletePending: False, Directory: False
2:33:52.4844948 PM  python.exe  2668    IRP_MJ_READ                     filename.csv    SUCCESS Offset: 117,318, Length: 1,024
2:33:52.4857179 PM  python.exe  2668    IRP_MJ_READ                     filename.csv    END OF FILE Offset: 118,342, Length: 4,096
2:33:52.4862472 PM  python.exe  2668    FASTIO_QUERY_INFORMATION        filename.csv    FAST IO DISALLOWED  Type: QueryStandardInformationFile
2:33:52.4862564 PM  python.exe  2668    IRP_MJ_QUERY_INFORMATION        filename.csv    SUCCESS Type: QueryStandardInformationFile, AllocationSize: 1,048,576, EndOfFile: 118,342, NumberOfLinks: 1, DeletePending: False, Directory: False
2:33:52.4867251 PM  python.exe  2668    IRP_MJ_READ                     filename.csv    SUCCESS Offset: 116,294, Length: 1,024
2:33:52.4873473 PM  python.exe  2668    IRP_MJ_READ                     filename.csv    SUCCESS Offset: 117,318, Length: 512
2:33:52.4878825 PM  python.exe  2668    FASTIO_QUERY_INFORMATION        filename.csv    FAST IO DISALLOWED  Type: QueryStandardInformationFile
2:33:52.4878917 PM  python.exe  2668    IRP_MJ_QUERY_INFORMATION        filename.csv    SUCCESS Type: QueryStandardInformationFile, AllocationSize: 1,048,576, EndOfFile: 118,342, NumberOfLinks: 1, DeletePending: False, Directory: False
2:33:52.4882153 PM  python.exe  2668    IRP_MJ_READ                     filename.csv    SUCCESS Offset: 115,270, Length: 1,024
2:33:52.4889601 PM  python.exe  2668    IRP_MJ_READ                     filename.csv    SUCCESS Offset: 116,294, Length: 512
2:33:52.4895164 PM  python.exe  2668    IRP_MJ_CLEANUP                  filename.csv    SUCCESS 
2:33:52.4895513 PM  python.exe  2668    IRP_MJ_CLOSE                    filename.csv    SUCCESS 

File open in another program (Errno 13 - Permission Denied):

1:12:42.0840918 PM  python.exe  5772    IRP_MJ_CREATE   filename.csv    SHARING VIOLATION   Desired Access: Generic Write, Read Attributes, Disposition: OpenIf, Options: Synchronous IO Non-Alert, Non-Directory File, Attributes: N, ShareMode: Read, Write, AllocationSize: 0

Misc error (not sure if this corresponds to a python error or not)

2:57:59.9371101 PM  python.exe  3584    IRP_MJ_CREATE   filename.log    BAD NETWORK PATH    Desired Access: Generic Read, Disposition: Open, Options: Synchronous IO Non-Alert, Non-Directory File, Attributes: N, ShareMode: Read, Write, AllocationSize: n/a

Would this show as an OSError 2?

Answer 1

If you're on Windows, use Process Monitor to record all of the file operations done by Python.exe processes. If you're on Mac OS X or Solaris, use DTrace. If you're on Linux, use strace.

Any of those tools will give you a detailed listing of all of the system calls being made, along with their return codes, allowing you to narrow down what's failing and perhaps why.

Answer 2

As a workaround you can wrap the code with exception handling (try/except)

A real solution is more complicated - synchronizing the scripts. This is what I would do: Write a broker script which listens to commands (UDP or SimpleXMLRPCServer, for example). Other scripts communicate with it. All operations on the shared file are done via the broker.

Answer 3

How a ftp share was created? how are organized 'open' files proceeding?

a) try to use 'with' according to pep 343.

if it not will work:

b) Try to make your own ftp share server with two methonds - read and
write. you can implement non-blocking reading algorithm.

Non-blocking algorithms are also safe for use in interrupt handlers: even though the preempted thread cannot be resumed, progress is still possible without it. In contrast, global data structures protected by mutual exclusion cannot safely be accessed in a handler, as the preempted thread may be the one holding the lock.