[sheepdog-users] Cluster snapshot out of memory

[Valerio Pachera]

It looks fixed to me.

Note: I was initially not able to start the cluster after the upgrade:

Sep 26 17:14:02   INFO [main] main(828) shutdown
Sep 26 17:35:25   INFO [main] md_add_disk(310) /mnt/sheep/dsk02, vdisk
nr 932, total disk 1
Sep 26 17:35:25  ERROR [main] do_epoch_log_read(78) invalid epoch 1 log
Sep 26 17:35:25  ERROR [main] main(759) failed to create sheepdog cluster
Sep 26 17:35:25  ERROR [main] crash_handler(492) sheep pid 28764
exited unexpectedly.
Sep 26 17:35:49   INFO [main] md_add_disk(310) /mnt/sheep/dsk02, vdisk
nr 932, total disk 1
Sep 26 17:35:49  ERROR [main] do_epoch_log_read(78) invalid epoch 1 log
Sep 26 17:35:49  ERROR [main] main(759) failed to create sheepdog cluster
Sep 26 17:35:49  ERROR [main] crash_handler(492) sheep pid 28781
exited unexpectedly.

I removed all data and restarted sheep, then it worked fine

Sep 26 17:38:06   INFO [main] md_add_disk(310) /mnt/sheep/dsk02, vdisk
nr 932, total disk 1
Sep 26 17:38:06   INFO [main] send_join_request(778) IPv4
ip:192.168.2.44 port:7000
Sep 26 17:38:06   INFO [main] check_host_env(420) Allowed open files
1024000, suggested 6144000
Sep 26 17:38:06   INFO [main] main(821) sheepdog daemon (version
0.7.0_131_gfd43e6f) started

top
top - 17:43:39 up 2 days,  5:17,  2 users,  load average: 0,90, 0,53, 0,26
Tasks:  88 total,   2 running,  86 sleeping,   0 stopped,   0 zombie
%Cpu(s): 10,3 us,  9,2 sy,  0,0 ni, 55,0 id, 22,1 wa,  0,0 hi,  3,4 si,  0,0 st
KiB Mem:   1929452 total,  1861252 used,    68200 free,     5136 buffers
KiB Swap:  7812092 total,    13604 used,  7798488 free,  1724968 cached
  PID USER      PR  NI  VIRT  RES  SHR S  %CPU %MEM    TIME+  COMMAND
28874 root      20   0 40888  12m  692 S  26,6  0,7   1:10.37 dog
28840 root      20   0 97644 8688 1956 S  13,3  0,5   0:35.76 sheep

Ram is used but it doesn't swap.

This is my script monitoring ram used by sheep (ps axv)

09/26/13 17:39:01 93190 2516
09/26/13 17:40:01 93314 4548
09/26/13 17:41:01 93314 4436
09/26/13 17:42:01 93314 4460
09/26/13 17:43:01 93314 4500
09/26/13 17:44:01 93314 4532
09/26/13 17:45:01 93314 4484

git branch -a
* devel
  master
  remotes/origin/HEAD -> origin/master
  remotes/origin/devel
  remotes/origin/gh-pages
  remotes/origin/master
  remotes/origin/option
  remotes/origin/reclaim
  remotes/origin/stable-0.6
  remotes/origin/stable-0.7

sheep -v
Sheepdog daemon version 0.7.0_131_gfd43e6f

git show -1
commit fd43e6f0e9450d916c916b908af8834c6aacae1c
Author: Liu Yuan <namei.unix at gmail.com>
Date:   Wed Sep 25 23:36:11 2013 +0800

    farm: fix memory leak of slice_read()

    This is yet another leaker of slice_read.

    Signed-off-by: Liu Yuan <namei.unix at gmail.com>

diff --git a/dog/farm/slice.c b/dog/farm/slice.c
index 79fa859..77db771 100644
--- a/dog/farm/slice.c
+++ b/dog/farm/slice.c
@@ -96,6 +96,7 @@ void *slice_read(const unsigned char *sha1, size_t *outsize)
                if (!sbuf)
                        goto err;
                strbuf_add(&buf, sbuf, size);
+               free(sbuf);
                *outsize += size;
        }