Assert error in stv_alloc(), stevedore.c line 144

[Reinis Rozitis]

Hello,
I am trying to switch from pure malloc (which works flawlesly, but there is 
a limit of just how much ram we can plug in a server) to some file storage 
on ssd but after a while (I imagine when it gets filled up) varnish (latest 
svn trunk) restarts:


Feb  5 23:02:15 is50 varnishd[12304]: Child (12305) not responding to CLI, 
killing it.
Feb  5 23:02:15 is50 varnishd[12304]: Child (12305) died signal=6
Feb  5 23:02:15 is50 varnishd[12304]: Child (12305) Panic message: Assert 
error in stv_alloc(), stevedore.c line 144:#012  Condition((st) != NULL) not 
true.#012thread = (cache-worker)#012ident = 
Linux,3.2.0-rc2-2-default,x86_64,-sfile,-sfile,-smalloc,-hcritbit,epoll#012Backtrace:#012 
0x4284a8: pan_ic+b8#012  0x43f9fd: STV_alloc+17d#012  0x41e057: 
vfp_nop_begin+97#012  0x41ea11: FetchBody+291#012  0x4164fd: 
CNT_Session+1c8d#012  0x429c28: wrk_do_cnt_sess+b8#012  0x42a059: 
wrk_thread_real+379#012  0x7f7dd3343a4f: _end+7f7dd2cd2937#012 
0x7f7dd30ad91d: _end+7f7dd2a3c805#012sp = 0x7f4b93e35008 {#012  fd = 228, id 
= 228, xid = 1707158194,#012  client = 10.0.0.2 33966,#012  step = 
STP_FETCH,#012  handling = deliver,#012  err_code = 200, err_reason = 
(null),#012  restarts = 0, esi_level = 0#012  ws = 0x7f4b93e35080 { #012 
id = "sess",#012    {s,f,r,e} = 
0x7f4b93e35cd8,+320,(nil),+65536},#012  },#012  http[req] = {#012    ws = 
0x7f4b93e35080[sess]#012      "GET",#012 
"/profile/544/700/v1328443824/544700.jpg",#012      "HTTP/1.1",#012 
"TE: deflate,gzip;q=0.3",#012      "Connection: TE, close",#012      "Host: 
i0.img.com",#012      "User-Agent: Mozilla/5.0 (Windows NT 6.1; rv:8.0) 
Gecko/20100101 Firefox/8.0",#012      "X-Forwarded-For: 
0.0.0.2",#012  },#012  worker = 0x7f4b2dbfebb0 {#012    ws = 0x7f4b2dbfed38 
{ #012      id = "wrk",#012      {s,f,r,e} = 
{0x7f4b2dbecb40,+4368,(nil),+65536},#012    },#012    http[bereq] = {#012 
ws = 0x7f4b2dbfed38[wrk]#012        "GET",#012 
"/profile/544/700/v1328443824/544700.jpg",#012        "HTTP/1.1",#012 
"Host: i0.img.com",#012        "User-Agent: Mozilla/5.0 (Windows NT 6.1; 
rv:8.0) Gecko/20100101 Firefox/8.0",#012        "X-Forwarded-For: 
10.0.0.2",#012        "X-Varnish: 1707158194",#012    },#012    http[beresp] 
= {#012      ws = 0x7f4b2dbfed38[wrk]#012        "HTTP/1.1",#012 
"200",#012        "OK",#012        "Date: Sun, 05 Feb 2012 21:02:11 
GMT",#012        "Content-Type: image/jpeg",#012        "Content-Length: 54
Feb  5 23:02:15 is50 varnishd[12304]: Child cleanup complete
Feb  5 23:02:15 is50 varnishd[12304]: child (21293) Started
Feb  5 23:02:15 is50 varnishd[12304]: Child (21293) said
Feb  5 23:02:15 is50 varnishd[12304]: Child (21293) said Child starts
Feb  5 23:02:15 is50 varnishd[12304]: Child (21293) said SMF.s0 mmap'ed 
107374182400 bytes of 107374182400
Feb  5 23:02:15 is50 varnishd[12304]: Child (21293) said SMF.s1 mmap'ed 
107374182400 bytes of 107374182400


Varnish is started with:

./varnishd -f default.vcl -s file,/mnt/ssd1/cache,100G -s 
file,/mnt/ssd2/cache,100G -h critbit -p connect_timeout 600 -p thread_pools 
8 -p thread_pool_min 100 -p thread_pool_max 5000 -thread_pool_add_delay=2 -p 
cli_timeout 50 -p session_linger 120 -p lru_interval 20


I have seen few similar bug reports:

https://www.varnish-cache.org/trac/ticket/927
https://www.varnish-cache.org/trac/ticket/759
https://www.varnish-cache.org/trac/ticket/503


So I wonder if this is something new or some of the old tickets to be 
reopened (or maybe the issue is solved somewhat different way)?


wbr
rr