From c2b1c50f06cc59b47c9c834617dff2aed7177a78 Mon Sep 17 00:00:00 2001 From: Ashish Pandey Date: Mon, 18 Mar 2019 12:54:54 +0530 Subject: [PATCH 164/169] cluster/ec: Fix handling of heal info cases without locks When we use heal info command, it takes lot of time as in some cases it takes lock on entries to find out if the entry actually needs heal or not. There are some cases where we can avoid these locks and can conclude if the entry needs heal or not. 1 - We do a lookup (without lock) on an entry, which we found in .glusterfs/indices/xattrop, and find that lock count is zero. Now if the file contains dirty bit set on all or any brick, we can say that this entry needs heal. 2 - If the lock count is one and dirty is greater than 1, then it also means that some fop had left the dirty bit set which made the dirty count of current fop (which has taken lock) more than one. At this point also we can definitely say that this entry needs heal. This patch is modifying code to take into consideration above two points. It is also changing code to not to call ec_heal_inspect if ec_heal_do was called from client side heal. Client side heal triggeres heal only when it is sure that it requires heal. [We have changed the code to not to call heal for lookup] Upstream patch - https://review.gluster.org/#/c/glusterfs/+/22372/ Fixes: bz#1716385 Change-Id: I7f09f0ecd12f65a353297aefd57026fd2bebdf9c Signed-off-by: Ashish Pandey Reviewed-on: https://code.engineering.redhat.com/gerrit/172579 Reviewed-by: Atin Mukherjee Tested-by: RHGS Build Bot --- xlators/cluster/ec/src/ec-heal.c | 42 ++++++++++++++++------------------------ 1 file changed, 17 insertions(+), 25 deletions(-) diff --git a/xlators/cluster/ec/src/ec-heal.c b/xlators/cluster/ec/src/ec-heal.c index 3aa04fb..2fa1f11 100644 --- a/xlators/cluster/ec/src/ec-heal.c +++ b/xlators/cluster/ec/src/ec-heal.c @@ -2541,13 +2541,15 @@ ec_heal_do(xlator_t *this, void *data, loc_t *loc, int32_t partial) /* Mount triggers heal only when it detects that it must need heal, shd * triggers heals periodically which need not be thorough*/ - ec_heal_inspect(frame, ec, loc->inode, up_subvols, _gf_false, - !ec->shd.iamshd, &need_heal); + if (ec->shd.iamshd) { + ec_heal_inspect(frame, ec, loc->inode, up_subvols, _gf_false, _gf_false, + &need_heal); - if (need_heal == EC_HEAL_NONEED) { - gf_msg(ec->xl->name, GF_LOG_DEBUG, 0, EC_MSG_HEAL_FAIL, - "Heal is not required for : %s ", uuid_utoa(loc->gfid)); - goto out; + if (need_heal == EC_HEAL_NONEED) { + gf_msg(ec->xl->name, GF_LOG_DEBUG, 0, EC_MSG_HEAL_FAIL, + "Heal is not required for : %s ", uuid_utoa(loc->gfid)); + goto out; + } } sources = alloca0(ec->nodes); healed_sinks = alloca0(ec->nodes); @@ -2902,7 +2904,7 @@ out: static int32_t _need_heal_calculate(ec_t *ec, uint64_t *dirty, unsigned char *sources, gf_boolean_t self_locked, int32_t lock_count, - ec_heal_need_t *need_heal) + ec_heal_need_t *need_heal, uint64_t *versions) { int i = 0; int source_count = 0; @@ -2912,7 +2914,7 @@ _need_heal_calculate(ec_t *ec, uint64_t *dirty, unsigned char *sources, *need_heal = EC_HEAL_NONEED; if (self_locked || lock_count == 0) { for (i = 0; i < ec->nodes; i++) { - if (dirty[i]) { + if (dirty[i] || (versions[i] != versions[0])) { *need_heal = EC_HEAL_MUST; goto out; } @@ -2928,6 +2930,9 @@ _need_heal_calculate(ec_t *ec, uint64_t *dirty, unsigned char *sources, *need_heal = EC_HEAL_MUST; goto out; } + if (dirty[i] != dirty[0] || (versions[i] != versions[0])) { + *need_heal = EC_HEAL_MAYBE; + } } } } else { @@ -2948,7 +2953,6 @@ ec_need_metadata_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies, unsigned char *healed_sinks = NULL; uint64_t *meta_versions = NULL; int ret = 0; - int i = 0; sources = alloca0(ec->nodes); healed_sinks = alloca0(ec->nodes); @@ -2961,15 +2965,7 @@ ec_need_metadata_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies, } ret = _need_heal_calculate(ec, dirty, sources, self_locked, lock_count, - need_heal); - if (ret == ec->nodes && *need_heal == EC_HEAL_NONEED) { - for (i = 1; i < ec->nodes; i++) { - if (meta_versions[i] != meta_versions[0]) { - *need_heal = EC_HEAL_MUST; - goto out; - } - } - } + need_heal, meta_versions); out: return ret; } @@ -3005,7 +3001,7 @@ ec_need_data_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies, } ret = _need_heal_calculate(ec, dirty, sources, self_locked, lock_count, - need_heal); + need_heal, data_versions); out: return ret; } @@ -3033,7 +3029,7 @@ ec_need_entry_heal(ec_t *ec, inode_t *inode, default_args_cbk_t *replies, } ret = _need_heal_calculate(ec, dirty, sources, self_locked, lock_count, - need_heal); + need_heal, data_versions); out: return ret; } @@ -3131,10 +3127,6 @@ ec_heal_inspect(call_frame_t *frame, ec_t *ec, inode_t *inode, need_heal: ret = ec_need_heal(ec, inode, replies, lock_count, self_locked, thorough, need_heal); - - if (!self_locked && *need_heal == EC_HEAL_MUST) { - *need_heal = EC_HEAL_MAYBE; - } out: cluster_replies_wipe(replies, ec->nodes); loc_wipe(&loc); @@ -3220,7 +3212,7 @@ ec_get_heal_info(xlator_t *this, loc_t *entry_loc, dict_t **dict_rsp) ret = ec_heal_inspect(frame, ec, loc.inode, up_subvols, _gf_false, _gf_false, &need_heal); - if (ret == ec->nodes && need_heal == EC_HEAL_NONEED) { + if (ret == ec->nodes && need_heal != EC_HEAL_MAYBE) { goto set_heal; } need_heal = EC_HEAL_NONEED; -- 1.8.3.1