处理文件的并发读/写

Question

I have a web service that is generating random errors and I think I've found the fault.

Basically, it is reading a config file as follows, then loaded into an XmlDocument:

var config = File.ReadAllText(filename);
xmlDoc.Load(config);

However, at a later time (maybe a second later) one of the config values is updated and the file saved

xmlDoc.Save(filename);

So I'm now experiencing more errors (unfortunately the original developer added an empty try block so I can't check just now) during the first READ operation and I think it's because it's trying to read the file just as another process spawned from IIS is at the .Save part. I don't know how File.ReadAllText works and whether it will fail on a write locked file.

What's the best solution to handle this to ensure reading will always work? the value being written is just a counter and if it fails it is ignored as it's not that important but would prefer it was written. I guess I could put it into a separate config file and live with the error but I'd rather it was just one file.

Thanks.

Answer 1

You can use a lock to make sure that a read is completed before a write and vice verser. As in:

using System;
using System.Threading;

class Program
{
    static readonly object _fileAccess = new object();

    static void Write()
    {
        // Obtain lock and write
        lock (_fileAccess)
        {
            // Write data to filename
            xmlDoc.Save(filename);
        }
    }

    static void Read()
    {
        // Obtain lock and read
        lock (_fileAccess)
        {
            // Read some data from filename
            xmlDoc.load(filename);
        }
    }

    static void Main()
    {       
        ThreadStart writeT = new ThreadStart(Write);
        new Thread(writeT).Start();

        ThreadStart readT = new ThreadStart(Read);
        new Thread(readT).Start();
    }
}

With the lock, the Read() must wait for the Write() to complete and Write() must wait for Read() to complete.

Answer 2

To answer your question about how File.ReadAllText() works, looking at the source, it uses a StreamReader internally which in turn uses a FileStream opened with FileAccess.Read and FileShare.Read , so that would prevent any other process from writing to the file (eg your XmlDocument.Save() ) until the ReadAllText completed.

Meanwhile, your XmlDocument.Save() eventually uses FileStream opened with FileAccess.Write and FileShare.Read , so it would allow the File.ReadAllText() as long as the Save started before the ReadAllText .

References: https://referencesource.microsoft.com/#mscorlib/system/io/streamreader.cs,a820588d8233a829

https://referencesource.microsoft.com/#System.Xml/System/Xml/Dom/XmlDocument.cs,1db4dba15523d588