Questions about the repair process

[httaotao]

Hello, there days, I am studing glusterfs, however, I am confused about the repair and I have some QA about repair in glusterfs.

1. I am confused about the glusterfs param cluster.data-self-heal-algorithm, what't the difference about param diffand full. For a file needed to be repaired, does it means that the diff is just to check the file checksum, and the full is to scan all the file to check?

2. For repair, how to control the repair speed ? For example, if there has a broken brick should be repair, I am confused that how to control the repair queue speed, because if there are too many files to repair, perhaps it would block the normal file read/write request.

3. if this file should be repaired, but then it also has the write request, then how to deal with this situation ? Does the write request should wait for the repair to complete or cancle repair to do write?

4. Do you decide whether to trigger the repair process only after the file is opened ? if a file never open again, does it has any times to do repair?

thanks.