Alon Bar-Lev has posted comments on this change.
Change subject: Avoiding automatically restart of sanlock service when starting
vdsmd
......................................................................
Patch Set 9:
(8 comments)
....................................................
File init/sysvinit/vdsmd.init.in
Line 61: initctl stop "${srv}" || : # stop fails when already
down
Line 62: initctl status "${srv}" | grep -q stop/waiting
Line 63: elif [ -x "/etc/init.d/${srv}" ]; then
Line 64: ! service "${srv}" status >/dev/null 2>&1 ||
Line 65: service "${srv}" stop
something in the indent is wrong
use:
service status && service stop
but we do want to have error of stop, so it should be:
if service status; then
service stop
fi
this way you will get rc=0 even if service status is failing
Line 66: else
Line 67: true
Line 68: fi
Line 69: fi
....................................................
File lib/vdsm/tool/configurator.py
Line 90: """
Line 91: Check if sanlock service requires a restart to reload the relevant
Line 92: supplementary groups.
Line 93: """
Line 94: retval = 0
always assume failure.
Line 95: sanlock_pid_file = "/var/run/sanlock/sanlock.pid"
Line 96: proc_status_path = "/proc/%s/status"
Line 97: proc_status_group_prefix = "Groups:\t"
Line 98:
Line 98:
Line 99: try:
Line 100: with open(sanlock_pid_file, "r") as f:
Line 101: sanlock_pid = f.readline().strip()
Line 102: sanlock_status = open(proc_status_path % sanlock_pid, "r")
leak?
Line 103: except IOError as e:
Line 104: if e.errno == os.errno.ENOENT:
Line 105: sys.stdout.write("sanlock service is not running\n")
Line 106: return retval # service is not running, returning
Line 100: with open(sanlock_pid_file, "r") as f:
Line 101: sanlock_pid = f.readline().strip()
Line 102: sanlock_status = open(proc_status_path % sanlock_pid, "r")
Line 103: except IOError as e:
Line 104: if e.errno == os.errno.ENOENT:
why not check if exist first then fail after?
if os.path.exists(...):
with open ... as
Line 105: sys.stdout.write("sanlock service is not running\n")
Line 106: return retval # service is not running, returning
Line 107: raise
Line 108:
Line 102: sanlock_status = open(proc_status_path % sanlock_pid, "r")
Line 103: except IOError as e:
Line 104: if e.errno == os.errno.ENOENT:
Line 105: sys.stdout.write("sanlock service is not running\n")
Line 106: return retval # service is not running, returning
please avoid return at middle.
Line 107: raise
Line 108:
Line 109: for status_line in sanlock_status:
Line 110: if status_line.startswith(proc_status_group_prefix):
Line 111: groups = [int(x) for x in
Line 112: status_line[len(proc_status_group_prefix):].
Line 113: strip().split(" ")]
Line 114: break
Line 115: else:
this is *UGLY*, close should be in finally anyway.
Line 116: sanlock_status.close()
Line 117: raise RuntimeError("Unable to find sanlock service groups")
Line 118:
Line 119: sanlock_status.close()
Line 117: raise RuntimeError("Unable to find sanlock service groups")
Line 118:
Line 119: sanlock_status.close()
Line 120: diskimage_gid = grp.getgrnam(DISKIMAGE_GROUP)[2]
Line 121: if diskimage_gid in groups:
not in... and in the success flow set retval = 0.
Now... question... why not throw RuntimeException with the error message each
time you have an error, and have proper positive flow?
Line 122: sys.stdout.write("sanlock service requires restart\n")
Line 123: else:
Line 124: sys.stdout.write("sanlock service is already configured\n")
Line 125: retval = 1
Line 137: service.service_stop("supervdsmd")
Line 138: service.service_stop("libvirtd")
Line 139:
Line 140: if not configure_libvirt(*args):
Line 141: sys.stderr.write("failed to run configure-libvirt\n")
and exit/throw with error? you cannot continue as usual.
Line 142: if sanlock_check_service(*args):
--
To view, visit http://gerrit.ovirt.org/19890
To unsubscribe, visit http://gerrit.ovirt.org/settings
Gerrit-MessageType: comment
Gerrit-Change-Id: I48658066f707632719df8d65799d26c7b239dcd5
Gerrit-PatchSet: 9
Gerrit-Project: vdsm
Gerrit-Branch: master
Gerrit-Owner: Yaniv Bronhaim <ybron...@redhat.com>
Gerrit-Reviewer: Alon Bar-Lev <alo...@redhat.com>
Gerrit-Reviewer: Dan Kenigsberg <dan...@redhat.com>
Gerrit-Reviewer: Yaniv Bronhaim <ybron...@redhat.com>
Gerrit-Reviewer: oVirt Jenkins CI Server
Gerrit-HasComments: Yes
_______________________________________________
vdsm-patches mailing list
vdsm-patches@lists.fedorahosted.org
https://lists.fedorahosted.org/mailman/listinfo/vdsm-patches
