[Varnish] #1119: Varnish 3.0.2 freezed, Pushing vcls failed:#012CLI communication error (hdr)

[Varnish]

#1119: Varnish 3.0.2 freezed, Pushing vcls failed:#012CLI communication error
(hdr)
-------------------------------------------------+-------------------------
 Reporter:  campisano                            |       Owner:
     Type:  defect                               |      Status:  closed
 Priority:  normal                               |   Milestone:
Component:  build                                |     Version:  3.0.2
 Severity:  blocker                              |  Resolution:  worksforme
 Keywords:  child died pushing vcls failed       |
  #012CLI communication error (hdr)              |
-------------------------------------------------+-------------------------

Comment (by lampe):

 Hi.

 Replying to [comment:7 phk]:
 > The connect_timeout has nothing to do with starting the child process.
 >
 > As I said, you can try to increase cli_timeout, if the problem is
 disk-i/o pileups.

 While I can understand the root cause for timeouts in master/client
 communications and fixed it by moving the shm log to tmpfs, I still think
 there's a bug or at least unexpected behaviour here.

 When the master fails to push the initial vcl to the child, it kills the
 child but does not try to restart it. The master process is left hanging
 useless without a child and requires a stop/start cycle.

 I can reproduce the problem with the shm log on HDD and large I/O, e.g.:

 while [ !`pidof varnishd |tr " " "\n" |wc -l` -eq 2 ]; do
   dd if=/dev/zero of=test bs=1M count=20480 conf=fdatasync
 done

 (varnish 3.0.3)

-- 
Ticket URL: <https://www.varnish-cache.org/trac/ticket/1119#comment:10>
Varnish <https://varnish-cache.org/>
The Varnish HTTP Accelerator