From 96575a86f55568e5e0a4d8d13aad1cf8a4aeb2b7 Mon Sep 17 00:00:00 2001
From: Ravishankar N <ravishankar@redhat.com>
Date: Fri, 30 Dec 2016 14:57:17 +0530
Subject: [PATCH 269/270] afr: Avoid resetting event_gen when brick is always
down
Backport of : http://review.gluster.org/16309
Problem:
__afr_set_in_flight_sb_status(), which resets event_gen to zero, is
called if failed_subvols[i] is non-zero for any brick. But failed_subvols[i]
is true even if the brick was down *before* the transaction started.
Hence say if 1 brick is down in a replica-3, every writev that comes
will trigger an inode refresh because of this resetting, as seen from
the no. of FSTATs in the profile info in the BZ.
Fix:
Reset event gen only if the brick was previously a valid read child and
the FOP failed on it the first time.
Also `s/afr_inode_read_subvol_reset/afr_inode_event_gen_reset` because
the function only resets event gen and not the data/metadata readable.
Change-Id: I2f6ccb792ad04b3b0d5cca5d5cb1a1d60e92046d
BUG: 1410025
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/94930
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
xlators/cluster/afr/src/afr-common.c | 24 +++++++++++++-----------
xlators/cluster/afr/src/afr-dir-write.c | 6 +++---
xlators/cluster/afr/src/afr.h | 6 +-----
3 files changed, 17 insertions(+), 19 deletions(-)
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index d16ddec..696e909 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -134,6 +134,7 @@ __afr_inode_ctx_get (xlator_t *this, inode_t *inode, afr_inode_ctx_t **ctx)
out:
return ret;
}
+
/*
* INODE CTX 64-bit VALUE FORMAT FOR SMALL (<= 16) SUBVOL COUNTS:
*
@@ -195,10 +196,7 @@ __afr_set_in_flight_sb_status (xlator_t *this, afr_local_t *local,
metadatamap_old = metadatamap = (val & 0x000000000000ffff);
datamap_old = datamap = (val & 0x00000000ffff0000) >> 16;
- /* Hard-code event to 0 since there is a failure and the inode
- * needs to be refreshed anyway.
- */
- event = 0;
+ event = (val & 0xffffffff00000000) >> 32;
if (txn_type == AFR_DATA_TRANSACTION)
tmp_map = datamap;
@@ -231,6 +229,8 @@ __afr_set_in_flight_sb_status (xlator_t *this, afr_local_t *local,
local->transaction.in_flight_sb = _gf_true;
metadatamap |= (1 << index);
}
+ if (metadatamap_old != metadatamap)
+ event = 0;
break;
case AFR_DATA_TRANSACTION:
@@ -240,10 +240,12 @@ __afr_set_in_flight_sb_status (xlator_t *this, afr_local_t *local,
local->transaction.in_flight_sb = _gf_true;
datamap |= (1 << index);
}
+ if (datamap_old != datamap)
+ event = 0;
break;
default:
- break;
+ break;
}
val = ((uint64_t) metadatamap) |
@@ -354,7 +356,7 @@ out:
}
int
-__afr_inode_read_subvol_reset_small (inode_t *inode, xlator_t *this)
+__afr_inode_event_gen_reset_small (inode_t *inode, xlator_t *this)
{
int ret = -1;
uint16_t datamap = 0;
@@ -455,7 +457,7 @@ out:
}
int
-__afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
+__afr_inode_event_gen_reset (inode_t *inode, xlator_t *this)
{
afr_private_t *priv = NULL;
int ret = -1;
@@ -463,7 +465,7 @@ __afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
priv = this->private;
if (priv->child_count <= 16)
- ret = __afr_inode_read_subvol_reset_small (inode, this);
+ ret = __afr_inode_event_gen_reset_small (inode, this);
else
ret = -1;
@@ -652,7 +654,7 @@ out:
}
int
-afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
+afr_inode_event_gen_reset (inode_t *inode, xlator_t *this)
{
int ret = -1;
@@ -660,7 +662,7 @@ afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
LOCK(&inode->lock);
{
- ret = __afr_inode_read_subvol_reset (inode, this);
+ ret = __afr_inode_event_gen_reset (inode, this);
}
UNLOCK(&inode->lock);
out:
@@ -2124,7 +2126,7 @@ afr_lookup_done (call_frame_t *frame, xlator_t *this)
if (afr_replies_interpret (frame, this, local->inode, NULL)) {
read_subvol = afr_read_subvol_decide (local->inode,
this, &args);
- afr_inode_read_subvol_reset (local->inode, this);
+ afr_inode_event_gen_reset (local->inode, this);
goto cant_interpret;
} else {
read_subvol = afr_data_subvol_get (local->inode, this,
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index 286a539..8e483c3 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -122,12 +122,12 @@ __afr_dir_write_finalize (call_frame_t *frame, xlator_t *this)
continue;
if (local->replies[i].op_ret < 0) {
if (local->inode)
- afr_inode_read_subvol_reset (local->inode, this);
+ afr_inode_event_gen_reset (local->inode, this);
if (local->parent)
- afr_inode_read_subvol_reset (local->parent,
+ afr_inode_event_gen_reset (local->parent,
this);
if (local->parent2)
- afr_inode_read_subvol_reset (local->parent2,
+ afr_inode_event_gen_reset (local->parent2,
this);
continue;
}
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 69d0584..c48507b 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -883,7 +883,7 @@ afr_inode_read_subvol_set (inode_t *inode, xlator_t *this,
int event_generation);
int
-afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this);
+afr_inode_event_gen_reset (inode_t *inode, xlator_t *this);
int
afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
@@ -907,10 +907,6 @@ afr_read_subvol_get (inode_t *inode, xlator_t *this, int *subvol_p,
afr_read_subvol_get(i, t, s, r, e, AFR_METADATA_TRANSACTION, a)
int
-afr_inode_ctx_reset_unreadable_subvol (inode_t *inode, xlator_t *this,
- int subvol_idx, int txn_type);
-
-int
afr_inode_refresh (call_frame_t *frame, xlator_t *this, inode_t *inode,
uuid_t gfid, afr_inode_refresh_cbk_t cbk);
--
2.9.3