search

utsaslab / crashmonkey

CrashMonkey: tools for testing file-system reliability (OSDI 18)
Apache License 2.0
189 stars 31 forks source link

Have a set of tests run against each file system #3

Closed vijay03 closed 6 years ago

vijay03 commented 7 years ago
  1. Test mmap() and file read()/write() at the same time. One process does file IO with read(), write(). Another process does mmap().
  2. Test crashes in the middle of rename
  3. Truncating a file, then creating another file
vijay03 commented 7 years ago
  1. Allocate resources such as inodes and data blocks until file system is 95% full. Then crash in the middle of workloads which allocate a resource, then deallocate the resource, repeatedly.
  2. Modify xfstests for crash testing
ashmrtn commented 7 years ago

For file systems that journal data, we should profile/test workloads where we fill the journal up so that it is forced to write it out

vijay03 commented 7 years ago

Agreed. In general I think our setup should always push the file system to the brink of resource exhaustion (90% space used, smallest journal possible etc) and then run our tests.