On Mon, Nov 26, 2018 at 8:24 AM Nir Soffer <nsoffer@redhat.com> wrote:
On Mon, Nov 12, 2018 at 7:03 AM Germano Veit Michel <germano@redhat.com> wrote:


On Fri, Nov 9, 2018 at 3:25 PM Germano Veit Michel <germano@redhat.com> wrote:


On Wed, Nov 7, 2018 at 9:34 PM Nir Soffer <nsoffer@redhat.com> wrote:
On Wed, Nov 7, 2018 at 6:27 AM Germano Veit Michel <germano@redhat.com> wrote:
Hi,

I have seen CI failing with some "random" errors which makes submitting a series of patches a bit annoying, as some patch always fails with some error, then next run another patch fails.

One of these problems is TestMount.testSymlinkMount. Looks like its getting EBUSY when doing a rmdir. So probably something still using it?

=================================== FAILURES ===================================
__________________________ TestMount.testSymlinkMount __________________________

args = (<storage.mount_test.TestMount testMethod=testSymlinkMount>,)
kwargs = {}

    @wraps(f)
    def wrapper(*args, **kwargs):
        if cond:
            raise SkipTest(reason)
>       return f(*args, **kwargs)

testValidation.py:274: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
testValidation.py:333: in wrapper
    return f(*args, **kwargs)
storage/mount_test.py:167: in testSymlinkMount
    m.umount()
/usr/lib64/python2.7/contextlib.py:35: in __exit__
    self.gen.throw(type, value, traceback)
testlib.py:218: in namedTemporaryDir
    shutil.rmtree(tmpDir)
/usr/lib64/python2.7/shutil.py:261: in rmtree
    rmtree(fullname, ignore_errors, onerror)
/usr/lib64/python2.7/shutil.py:270: in rmtree
    onerror(os.rmdir, path, sys.exc_info())
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

path = '/var/tmp/tmpOYMiHx/mountpoint', ignore_errors = False
onerror = <function onerror at 0x7fd68afc9938>

    def rmtree(path, ignore_errors=False, onerror=None):
        """Recursively delete a directory tree.
    
        If ignore_errors is set, errors are ignored; otherwise, if onerror
        is set, it is called to handle the error with arguments (func,
        path, exc_info) where func is os.listdir, os.remove, or os.rmdir;
        path is the argument to that function that caused it to fail; and
        exc_info is a tuple returned by sys.exc_info().  If ignore_errors
        is false and onerror is None, an exception is raised.
    
        """
        if ignore_errors:
            def onerror(*args):
                pass
        elif onerror is None:
            def onerror(*args):
                raise
        try:
            if os.path.islink(path):
                # symlinks to directories are forbidden, see bug #1669
                raise OSError("Cannot call rmtree on a symbolic link")
        except OSError:
            onerror(os.path.islink, path, sys.exc_info())
            # can't continue even if onerror hook returns
            return
        names = []
        try:
            names = os.listdir(path)
        except os.error, err:
            onerror(os.listdir, path, sys.exc_info())
        for name in names:
            fullname = os.path.join(path, name)
            try:
                mode = os.lstat(fullname).st_mode
            except os.error:
                mode = 0
            if stat.S_ISDIR(mode):
                rmtree(fullname, ignore_errors, onerror)
            else:
                try:
                    os.remove(fullname)
                except os.error, err:
                    onerror(os.remove, fullname, sys.exc_info())
        try:
>           os.rmdir(path)
E           OSError: [Errno 16] Device or resource busy: '/var/tmp/tmpOYMiHx/mountpoint'

/usr/lib64/python2.7/shutil.py:268: OSError
----------------------------- Captured stderr call -----------------------------
2018-11-07 00:41:55,518 DEBUG (MainThread) [storage.Misc.excCmd] /usr/bin/taskset --cpu-list 0-1 /sbin/mkfs.ext2 -F /var/tmp/tmpOYMiHx/backing.img (cwd None) (commands:68)
2018-11-07 00:41:55,561 DEBUG (MainThread) [storage.Misc.excCmd] SUCCESS: <err> = 'mke2fs 1.44.2 (14-May-2018)\n'; <rc> = 0 (commands:89)
2018-11-07 00:41:55,562 INFO  (MainThread) [storage.Mount] mounting /var/tmp/tmpOYMiHx/link_to_image at /var/tmp/tmpOYMiHx/mountpoint (mount:204)
2018-11-07 00:41:55,563 DEBUG (MainThread) [root] /usr/bin/taskset --cpu-list 0-1 /usr/bin/mount -o loop /var/tmp/tmpOYMiHx/link_to_image /var/tmp/tmpOYMiHx/mountpoint (cwd None) (commands:68)
2018-11-07 00:41:55,598 DEBUG (MainThread) [root] SUCCESS: <err> = ''; <rc> = 0 (commands:89)
2018-11-07 00:41:55,599 DEBUG (MainThread) [storage.Mount] /var/tmp/tmpOYMiHx/mountpoint mounted: 0.03 seconds (utils:454)
2018-11-07 00:41:55,599 DEBUG (MainThread) [root] /usr/bin/taskset --cpu-list 0-1 /sbin/udevadm settle --timeout=5 (cwd None) (commands:68)
2018-11-07 00:41:55,624 DEBUG (MainThread) [root] SUCCESS: <err> = ''; <rc> = 0 (commands:89)
2018-11-07 00:41:55,625 DEBUG (MainThread) [storage.Mount] Waiting for udev mount events: 0.03 seconds (utils:454)
2018-11-07 00:41:55,628 INFO  (MainThread) [storage.Mount] unmounting /var/tmp/tmpOYMiHx/mountpoint (mount:212)
2018-11-07 00:41:55,629 DEBUG (MainThread) [root] /usr/bin/taskset --cpu-list 0-1 /usr/bin/umount /var/tmp/tmpOYMiHx/mountpoint (cwd None) (commands:68)
2018-11-07 00:41:55,658 DEBUG (MainThread) [root] FAILED: <err> = 'umount: /var/tmp/tmpOYMiHx/mountpoint: target is busy.\n'; <rc> = 32 (commands:89)

The test is doing:

163             m.mount(mntOpts="loop")
164             try:
165                 self.assertTrue(m.isMounted())
166             finally:
167                 m.umount()

So nothing can use this mount - except systemd, creating a mount unit for every mount
created by "mount".

If you can reproduce this with calling "mount" and "umount" in a loop, we can open
a systemd bug for this.

Adding little sleep before unmounting will probably avoid this issue. In production we
never unmount write after mount.

A possible fix is to use m.umount(force=True, lazy=True) - this is how we unmount
file based storage mounts in lib/vdsm/storage/storageServer.py:

201     def disconnect(self):
202         self._mount.umount(True, True)

Nir 

Thanks Nir!

I'll see if I can reproduce it next week. I have only seen this one on FC28, so it could be a FC28 bug. 
Yesterday the rebase CI also failed on 2 patches, but the errors were different and I did not have time to investigate. 

Hit it again, but now on testLoopMount() instead of testSymlinkMount().

Tried reproducing it on F29 (not 28), did not hit it.

#!/bin/bash

truncate -s 100M /tmp/a
mkfs.ext2 /tmp/a
mkdir /tmp/b
while [ True ]
do
    mount -o loop /tmp/a /tmp/b
    umount /tmp/b
done

Do you want me to submit a patch with m.umount(force=True, lazy=True) for all those tests? Would it be correct to do this?

I reproduce these issue today in the CI, and look at git history.

We had force=True in the past, but it was removed in:

commit c8a216677e82aa61aab0c98a1021265121f8db57
Author: Nir Soffer <nsoffer@redhat.com>
Date:   Sun Apr 24 01:05:33 2016 +0300

    mount: Remove pointless force=True usage
    
    Using force=True when unmounting loop device is pointless - this option
    will cause umount(8) to call the umount(2) syscall with the MNT_FORCE
    flag, which is useful only for NFS mounts.
    
    Change-Id: I31ff97d882f292f9c6325903ed9072b7146f682e
    Signed-off-by: Nir Soffer <nsoffer@redhat.com>
    Reviewed-by: Dan Kenigsberg <danken@redhat.com>
    Reviewed-by: Allon Mureinik <amureini@redhat.com>
    Continuous-Integration: Jenkins CI

We fixed similar issue later in:

commit d158b1a6f54ba67d81cb163593e3c330c77d36fd
Author: Nir Soffer <nsoffer@redhat.com>
Date:   Sun Apr 24 22:36:47 2016 +0300

    mount: Wait for systemd temporary mount unit
    
    Systemd is monitoring mountpoints created during runtime by creating a
    temporary mount units. For example:
    
        # systemctl status /tmp/mount-1.DUNxySxI2R/wrk
        ● tmp-mount\x2d1.DUNxySxI2R-wrk.mount - /tmp/mount-1.DUNxySxI2R/wrk
           Loaded: loaded (/proc/self/mountinfo)
           Active: active (mounted) since Sun 2016-04-24 22:37:55 IDT; 14ms ago
            Where: /tmp/mount-1.DUNxySxI2R/wrk
             What: /dev/loop2
    
    Mounting and unmounting quickly as we do in the tests, or when creating
    a floppy image may confuse systemd, and lead to various errors:
    
    - mountpoint is not mounted when trying to access files or unmount
    - mountpoint is busy when trying to unmount
    
    These are probably systemd or mount bugs. Until these issues are
    resolve, we need to find a workaround to avoid failures in the tests or
    when creating a floppy images.
    
    This patch add udevadm.settle() call after successful mount and umount
    operations.
    
    Change-Id: I9b0318eb534f781f127882983527e1a221820478
    Signed-off-by: Nir Soffer <nsoffer@redhat.com>
    Reviewed-by: Dan Kenigsberg <danken@redhat.com>
    Continuous-Integration: Jenkins CI
    Reviewed-by: Francesco Romani <fromani@redhat.com>

Based on this, I posted

Please try to rebase your patches on top of this, and report if it fixes
this issue.

Nir

Hi Nir,

It makes sense. So I did rebase my patches on top of yours, and it seems to have fixed those random errors. It used to get at least 2 failures out of 12 patches. After the rebase I hit 0 of those errors.

Still, I hit 4 of these on the latest round, but its not the same problem. So it looks like your patch did the job.

ERROR: Command failed: 
 # /usr/sbin/groupadd -g 1000 mock

Thanks!
Germano