Description of problem:
vdsm needs lstat() to implement os.path.islink(). Since lstat() may block if the
link is on shared storage, we need to implement it in ioprocess.
How to test the new feature:
1. Create a symlink
$ ln -s /no/such/file test-link
2. Print symlink stats
$ python -c 'import ioprocess; print ioprocess.IOProcess().lstat("test-link")'
StatResult(st_mode=41471, st_ino=8521781, st_dev=64772, st_nlink=1, st_uid=1000, st_gid=1000, st_size=13, st_atime=1523021206.0, st_mtime=1523021206.0, st_ctime=1523021206.0, st_blocks=0)
Tal, we need this in 4.2.z. Can you add the necessary flags?
Nir, both patches are merged. Should this be MODIFIED? Are we going to have another ioprocess build for 4.2?
We is the functional impact of this RFE?
Allon: yes, it is modified, ioprocess bugzilla integration is broken for some
reason. I will build a new ioprocess for 4.2.z.
Yaniv: see the vdsm bug depending on this.
I haven't seen an ioprocess build for 4.2.4 including this fix. Did you build?
ioprocess was not build and doubt if it will make it to 4.2.4, it's a minor issue so I'm deferring to 4.2.5
Nir, is there a specific flow we should test here?
Moving back to POST since patches have been merged to master but not backported to ioprocess-1.0 branch for 4.2.5
(In reply to Sandro Bonazzola from comment #5)
> I haven't seen an ioprocess build for 4.2.4 including this fix. Did you
Eyal is working on the build for 4.2.5.
Re adding need info from comment #7
(In reply to Elad from comment #7)
> Nir, is there a specific flow we should test here?
You can try to reproduce bug 1560460, probably by adding a symlink in storage
manually. I'm not sure it will be easy to reproduce and I never tried it.
The new lstat() api has automated tests so I don't see a need to waste time on
checking the new api manually.
The new api is used deep inside vdsm so many flows may be affected, so the standard
regression tests is the best way to verify this.
Thanks Nir, we will verify based on our regression tests results.
This is available in ioprocess 1.1.2 for some time.
Haven't seen any related issue in the last regression automation executions.