[Varnish] #1082: Deadlock?

[Varnish]

#1082: Deadlock?
--------------------+-------------------------------------------------------
 Reporter:  philip  |        Type:  defect  
   Status:  new     |    Priority:  normal  
Milestone:          |   Component:  varnishd
  Version:  3.0.0   |    Severity:  normal  
 Keywords:          |  
--------------------+-------------------------------------------------------
 Hi,

 I just rolled out varnish into production with the latest stable 3
 release. After serving lots of requests and peaking at 600Mbps I got
 something that looked like a deadlook. The server was not responding to
 requests anymore. I routed all the requests away from the varnish cache to
 the backend.

 After no requests reaching the varnish server anymore I checked the system
 usage with "atop". The HDD was under heavy use:

 {{{

 DSK |         sda | busy    104% | read    4044 | write      6 | avio    2
 ms |
 DSK |         sdb | busy    104% | read   13604 | write      7 | avio    0
 ms |
 NET | transport   | tcpi    3321 | tcpo    6472 | udpi       0 | udpo
 7 |
 NET | network     | ipi     3333 | ipo     2291 | ipfrw      0 | deliv
 3333 |
 NET | eth0     0% | pcki    3333 | pcko    6739 | si  172 Kbps | so 8029
 Kbps |

 }}}


 This server *only* runs varnish. I have attached the /etc/default/varnish
 and /etc/varnish/default.vcl files.

-- 
Ticket URL: <https://www.varnish-cache.org/trac/ticket/1082>
Varnish <https://varnish-cache.org/>
The Varnish HTTP Accelerator