From 70d5e4931e07f82ceed53d2934167b1ebd74f2e9 Mon Sep 17 00:00:00 2001

From: Mohit Agrawal <moagrawa@redhat.com>

Date: Thu, 22 Jun 2017 16:57:04 +0530

Subject: [PATCH 531/539] glusterd: brick process fails to restart after

 gluster pod failure

Problem: In container environment sometime after delete gluster pod

         and created new gluster pod brick process doesn't seem

         to come up.

Solution: On the basis of logs it seems glusterd is trying to attach

          with non glusterfs process.Change the code of function

          glusterd_get_sock_from_brick_pid to fetch socketpath from argument

          of running brick process.

> BUG: 1464072

> Change-Id: Ida6af00066341b683bbb4440d7a0d8042581656a

> Signed-off-by: Mohit Agrawal <moagrawa@redhat.com>

> Reviewed-on: https://review.gluster.org/17601

> Smoke: Gluster Build System <jenkins@build.gluster.org>

> CentOS-regression: Gluster Build System <jenkins@build.gluster.org>

> NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>

> Reviewed-by: Atin Mukherjee <amukherj@redhat.com>

> (Cherry picked from commit b71059960f8c67d9a058244d2a1c748be4fe1323)

BUG: 1463221

Change-Id: I9ac21ee7150fb2f17157fab6dc6dde72f329d80e

Signed-off-by: Mohit Agrawal <moagrawa@redhat.com>

Reviewed-on: https://code.engineering.redhat.com/gerrit/110184

Reviewed-by: Atin Mukherjee <amukherj@redhat.com>

---

 xlators/mgmt/glusterd/src/glusterd-utils.c | 41 ++++++++++++++++++++++--------

 1 file changed, 31 insertions(+), 10 deletions(-)

diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c

index fcb4340..18249ea 100644

--- a/xlators/mgmt/glusterd/src/glusterd-utils.c

+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c

@@ -5334,10 +5334,11 @@ find_compatible_brick (glusterd_conf_t *conf,

 }

 /* Below function is use to populate sockpath based on passed pid

-   value as a argument after check the value from proc

+   value as a argument after check the value from proc and also

+   check if passed pid is match with running  glusterfs process

 */

-void

+int

 glusterd_get_sock_from_brick_pid (int pid, char *sockpath, size_t len)

 {

         char fname[128] = {0,};

@@ -5350,6 +5351,7 @@ glusterd_get_sock_from_brick_pid (int pid, char *sockpath, size_t len)

         char   *brptr   = NULL;

         char tmpsockpath[PATH_MAX] = {0,};

         size_t blen    = 0;

+        int    ret     = -1;

         this = THIS;

         GF_ASSERT (this);

@@ -5359,7 +5361,7 @@ glusterd_get_sock_from_brick_pid (int pid, char *sockpath, size_t len)

         if (sys_access (fname , R_OK) != 0) {

                 gf_log (this->name, GF_LOG_ERROR,

                          "brick process %d is not running", pid);

-                return;

+                return ret;

         }

         fd = open(fname, O_RDONLY);

@@ -5369,7 +5371,7 @@ glusterd_get_sock_from_brick_pid (int pid, char *sockpath, size_t len)

                 gf_log (this->name, GF_LOG_ERROR,

                          "open failed %s to open a file %s", strerror (errno),

                                                               fname);

-                return;

+                return ret;

         }

         /* convert cmdline to single string */

@@ -5388,10 +5390,18 @@ glusterd_get_sock_from_brick_pid (int pid, char *sockpath, size_t len)

         cmdline[j] = '\0';

         if (fd)

                 sys_close(fd);

+        if (!strstr (cmdline, "glusterfs"))

+                return ret;

-        ptr =   strstr(cmdline, "-S ");

-        ptr =   strchr(ptr, '/');

+        ptr = strstr(cmdline, "-S ");

+        if (!ptr)

+                return ret;

+        ptr = strchr(ptr, '/');

+        if (!ptr)

+                return ret;

         brptr = strstr(ptr, "--brick-name");

+        if (!brptr)

+                return ret;

         i = 0;

         while (ptr < brptr) {

@@ -5402,8 +5412,10 @@ glusterd_get_sock_from_brick_pid (int pid, char *sockpath, size_t len)

         if (tmpsockpath[0]) {

                 strncpy (sockpath, tmpsockpath , i);

+                ret = 0;

         }

+        return ret;

 }

@@ -5466,22 +5478,31 @@ glusterd_brick_start (glusterd_volinfo_t *volinfo,

                          * same port (on another brick) and re-use that.

                          * TBD: re-use RPC connection across bricks

                          */

-                        if (is_brick_mx_enabled ())

-                                glusterd_get_sock_from_brick_pid (pid, socketpath,

-                                                                  sizeof(socketpath));

-                        else

+                        if (is_brick_mx_enabled ()) {

+                                ret = glusterd_get_sock_from_brick_pid (pid, socketpath,

+                                                                        sizeof(socketpath));

+                                if (ret) {

+                                        gf_log (this->name, GF_LOG_DEBUG,

+                                                "Either pid %d is not running or is not match"

+                                                " with any running brick process ", pid);

+                                        goto run;

+                                }

+                        } else {

                                 glusterd_set_brick_socket_filepath (volinfo, brickinfo,

                                                                     socketpath,

                                                                     sizeof (socketpath));

+                        }

                         gf_log (this->name, GF_LOG_DEBUG,

                                 "Using %s as sockfile for brick %s of volume %s ",

                                 socketpath, brickinfo->path, volinfo->volname);

+

                         (void) glusterd_brick_connect (volinfo, brickinfo,

                                         socketpath);

                 }

                 return 0;

         }

+run:

         ret = _mk_rundir_p (volinfo);

         if (ret)

                 goto out;

-- 

1.8.3.1