[ anon ]
2aab2800 65372 65372 65372 rw---[ anon ]
2aab2bfd7000 164 0 0 -[ anon ]
2aab2c00 61408 61152 61152 rw---[ anon ]
2aab3000 63468 63468 63468 rw---[ anon ]
2aab33dfb0002068 0 0 -[ anon ]
2aab3400 47816 47540 47540 rw---[ anon ]
2aab3800 33204 14516 14516 rw---[ anon ]
2aab3a06d000 32332 0 0 -[ anon ]
2ae3ff02d000 4 4 4 rw---[ anon ]
2ae3ff03e000 276 276 276 rw---[ anon ]
7fff8fa4a000 84 20 20 rw---[ stack ]
ff608192 0 0 -[ anon ]
-- -- --
total kB 3161504 3029568 3027536
strace -fp:
Process 2248 attached with 5 threads - interrupt to quit
[pid 2252] epoll_wait(7,
[pid 2251] clock_gettime(CLOCK_REALTIME,
[pid 2250] futex(0x2aaab104c088, FUTEX_WAIT_PRIVATE, 2, NULL
[pid 2249] futex(0x2ae3ff047084, FUTEX_WAIT_PRIVATE, 4239917443, NULL
[pid 2248] rt_sigsuspend([]
[pid 2251] <... clock_gettime resumed> {1275976570, 97551000}) = 0
[pid 2251] futex(0x2ae3ff048074, FUTEX_WAIT_PRIVATE, 93569205, {0, 301251000})
= -1 ETIMEDOUT (Connection timed out)
[pid 2251] futex(0x2ae3ff048020, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 2251] clock_gettime(CLOCK_REALTIME, {1275976570, 400051000}) = 0
[pid 2251] futex(0x2ae3ff048074, FUTEX_WAIT_PRIVATE, 93569207, {0, 252521000})
= -1 ETIMEDOUT (Connection timed out)
[pid 2251] futex(0x2ae3ff048020, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 2251] clock_gettime(CLOCK_REALTIME, {1275976570, 654023000}) = 0
[pid 2251] futex(0x2ae3ff048074, FUTEX_WAIT_PRIVATE, 93569209, {0, 75751000})
= -1 ETIMEDOUT (Connection timed out)
[pid 2251] futex(0x2ae3ff048020, FUTEX_WAKE_PRIVATE, 1) = 0
[pid 2251] clock_gettime(CLOCK_REALTIME, {1275976570, 731031000}) = 0
[pid 2251] futex(0x2ae3ff048074, FUTEX_WAIT_PRIVATE, 93569211, {0, 155742000})
= -1 ETIMEDOUT (Connection timed out)
[pid 2251] futex(0x2ae3ff048020, FUTEX_WAKE_PRIVATE, 1) = 0
>From what I can understand the threads are hanging waiting for lock and
nothing happens afterwards.
Without running 'rndc flush' the bind will eventually reach 4G and crash
with some other error which I currently don't have.
Up to now we tried different max-cache settings and threaded/non-threaded
compilations without much difference.
In all situations the named is 64-bit executable.
The problem never happens with bind 9.4.3-P5 that we run (nor with older
version of 9.4), so it seems that from 9.6 (maybe even 9.5) the memory
management changed. I also tried tests with 9.7.0-P1/P2 with same outcome.
Any help on the issue will be greatly appreciated. I'm open to any suggestions.
Thanks in advance.
Stas Pirogov
013 Netvision
___
bind-users mailing list
bind-users@lists.isc.org
https://lists.isc.org/mailman/listinfo/bind-users