|
 |
d1681e |
From 029fbbdaa7c4ddcc2479f507345a5c3ab1035313 Mon Sep 17 00:00:00 2001
|
|
|
d1681e |
From: Ravishankar N <ravishankar@redhat.com>
|
|
|
d1681e |
Date: Mon, 2 Jul 2018 16:05:39 +0530
|
|
|
d1681e |
Subject: [PATCH 306/325] glusterfsd: Do not process GLUSTERD_BRICK_XLATOR_OP
|
|
|
d1681e |
if graph is not ready
|
|
|
d1681e |
|
|
|
d1681e |
Patch in upstream master: https://review.gluster.org/#/c/20435/
|
|
|
d1681e |
Patch in release-3.12: https://review.gluster.org/#/c/20436/
|
|
|
d1681e |
|
|
|
d1681e |
Problem:
|
|
|
d1681e |
If glustershd gets restarted by glusterd due to node reboot/volume start force/
|
|
|
d1681e |
or any thing that changes shd graph (add/remove brick), and index heal
|
|
|
d1681e |
is launched via CLI, there can be a chance that shd receives this IPC
|
|
|
d1681e |
before the graph is fully active. Thus when it accesses
|
|
|
d1681e |
glusterfsd_ctx->active, it crashes.
|
|
|
d1681e |
|
|
|
d1681e |
Fix:
|
|
|
d1681e |
Since glusterd does not really wait for the daemons it spawned to be
|
|
|
d1681e |
fully initialized and can send the request as soon as rpc initialization has
|
|
|
d1681e |
succeeded, we just handle it at shd. If glusterfs_graph_activate() is
|
|
|
d1681e |
not yet done in shd but glusterd sends GD_OP_HEAL_VOLUME to shd,
|
|
|
d1681e |
we fail the request.
|
|
|
d1681e |
|
|
|
d1681e |
Change-Id: If6cc07bc5455c4ba03458a36c28b63664496b17d
|
|
|
d1681e |
BUG: 1593865
|
|
|
d1681e |
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
d1681e |
Reviewed-on: https://code.engineering.redhat.com/gerrit/143097
|
|
|
d1681e |
Tested-by: RHGS Build Bot <nigelb@redhat.com>
|
|
|
d1681e |
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
|
|
|
d1681e |
---
|
|
|
d1681e |
glusterfsd/src/glusterfsd-messages.h | 4 +++-
|
|
|
d1681e |
glusterfsd/src/glusterfsd-mgmt.c | 6 ++++++
|
|
|
d1681e |
2 files changed, 9 insertions(+), 1 deletion(-)
|
|
|
d1681e |
|
|
|
d1681e |
diff --git a/glusterfsd/src/glusterfsd-messages.h b/glusterfsd/src/glusterfsd-messages.h
|
|
|
d1681e |
index e9c28f7..e38a88b 100644
|
|
|
d1681e |
--- a/glusterfsd/src/glusterfsd-messages.h
|
|
|
d1681e |
+++ b/glusterfsd/src/glusterfsd-messages.h
|
|
|
d1681e |
@@ -36,7 +36,7 @@
|
|
|
d1681e |
*/
|
|
|
d1681e |
|
|
|
d1681e |
#define GLFS_COMP_BASE GLFS_MSGID_COMP_GLUSTERFSD
|
|
|
d1681e |
-#define GLFS_NUM_MESSAGES 37
|
|
|
d1681e |
+#define GLFS_NUM_MESSAGES 38
|
|
|
d1681e |
#define GLFS_MSGID_END (GLFS_COMP_BASE + GLFS_NUM_MESSAGES + 1)
|
|
|
d1681e |
/* Messaged with message IDs */
|
|
|
d1681e |
#define glfs_msg_start_x GLFS_COMP_BASE, "Invalid: Start of messages"
|
|
|
d1681e |
@@ -109,6 +109,8 @@
|
|
|
d1681e |
#define glusterfsd_msg_36 (GLFS_COMP_BASE + 36), "problem in xlator " \
|
|
|
d1681e |
" loading."
|
|
|
d1681e |
#define glusterfsd_msg_37 (GLFS_COMP_BASE + 37), "failed to get dict value"
|
|
|
d1681e |
+#define glusterfsd_msg_38 (GLFS_COMP_BASE + 38), "Not processing brick-op no."\
|
|
|
d1681e |
+ " %d since volume graph is not yet active."
|
|
|
d1681e |
|
|
|
d1681e |
/*------------*/
|
|
|
d1681e |
#define glfs_msg_end_x GLFS_MSGID_END, "Invalid: End of messages"
|
|
|
d1681e |
diff --git a/glusterfsd/src/glusterfsd-mgmt.c b/glusterfsd/src/glusterfsd-mgmt.c
|
|
|
d1681e |
index 665b62c..2167241 100644
|
|
|
d1681e |
--- a/glusterfsd/src/glusterfsd-mgmt.c
|
|
|
d1681e |
+++ b/glusterfsd/src/glusterfsd-mgmt.c
|
|
|
d1681e |
@@ -790,6 +790,12 @@ glusterfs_handle_translator_op (rpcsvc_request_t *req)
|
|
|
d1681e |
|
|
|
d1681e |
ctx = glusterfsd_ctx;
|
|
|
d1681e |
active = ctx->active;
|
|
|
d1681e |
+ if (!active) {
|
|
|
d1681e |
+ ret = -1;
|
|
|
d1681e |
+ gf_msg (this->name, GF_LOG_ERROR, EAGAIN, glusterfsd_msg_38,
|
|
|
d1681e |
+ xlator_req.op);
|
|
|
d1681e |
+ goto out;
|
|
|
d1681e |
+ }
|
|
|
d1681e |
any = active->first;
|
|
|
d1681e |
input = dict_new ();
|
|
|
d1681e |
ret = dict_unserialize (xlator_req.input.input_val,
|
|
|
d1681e |
--
|
|
|
d1681e |
1.8.3.1
|
|
|
d1681e |
|