diff options
author | Xavi Hernandez <xhernandez@redhat.com> | 2019-04-12 17:54:44 +0200 |
---|---|---|
committer | Pranith Kumar Karampuri <pkarampu@redhat.com> | 2019-04-23 11:28:58 +0000 |
commit | 4f0db1373be93e05b6fc451d2f07514704d3c4ca (patch) | |
tree | 82f4d82fbe914bcee2e900a87281a681b6cd40fc /xlators/cluster/ec/src/ec-common.c | |
parent | 132f03a0b77980b25a8b499ad1fe1e4ab2aee40e (diff) |
cluster/ec: fix fd reopen
Currently EC tries to reopen fd's that have been opened while a brick
was down. This is done as part of regular write operations, just after
having acquired the locks, and it's sent as a sub-fop of the main write
fop.
There were two problems:
1. The reopen was attempted on all UP bricks, even if a previous lock
didn't succeed. This is incorrect because most probably the open will
fail.
2. If reopen is sent and fails, the error is propagated to the main
operation, causing it to fail when it shouldn't.
To fix this, we only attempt reopens on bricks where the current fop
owns a lock, and we prevent any error to be propagated to the main
fop.
To implement this behaviour an argument used to indicate the minimum
number of required answers has overloaded to also include some flags. To
make the change consistent, it has been necessary to rename the
argument, which means that a lot of files have been changed. However
there are no functional changes.
This change has also uncovered a problem in discard code, which didn't
correctely process requests of small sizes because no real discard fop
was being processed, only a write of 0's on some region. In this case
some fields of the fop remained uninitialized or with incorrect values.
To fix this, a new function has been created to simulate success on a
fop and it's used in the discard case.
Thanks to Pranith for providing a test script that has also detected an
issue in this patch. This patch includes a small modification of this
script to force data to be written into bricks before stopping them.
Change-Id: If272343873369186c2fb8f43c1d9c52c3ea304ec
Fixes: bz#1699866
Signed-off-by: Xavi Hernandez <xhernandez@redhat.com>
Diffstat (limited to 'xlators/cluster/ec/src/ec-common.c')
-rw-r--r-- | xlators/cluster/ec/src/ec-common.c | 73 |
1 files changed, 53 insertions, 20 deletions
diff --git a/xlators/cluster/ec/src/ec-common.c b/xlators/cluster/ec/src/ec-common.c index 518368073e8..1454ae23b23 100644 --- a/xlators/cluster/ec/src/ec-common.c +++ b/xlators/cluster/ec/src/ec-common.c @@ -44,16 +44,16 @@ ec_update_fd_status(fd_t *fd, xlator_t *xl, int idx, int32_t ret_status) UNLOCK(&fd->lock); } -static int -ec_fd_ctx_need_open(fd_t *fd, xlator_t *this, uintptr_t *need_open) +static uintptr_t +ec_fd_ctx_need_open(fd_t *fd, xlator_t *this, uintptr_t mask) { int i = 0; int count = 0; ec_t *ec = NULL; ec_fd_t *fd_ctx = NULL; + uintptr_t need_open = 0; ec = this->private; - *need_open = 0; fd_ctx = ec_fd_get(fd, this); if (!fd_ctx) @@ -63,9 +63,9 @@ ec_fd_ctx_need_open(fd_t *fd, xlator_t *this, uintptr_t *need_open) { for (i = 0; i < ec->nodes; i++) { if ((fd_ctx->fd_status[i] == EC_FD_NOT_OPENED) && - (ec->xl_up & (1 << i))) { + ((ec->xl_up & (1 << i)) != 0) && ((mask & (1 << i)) != 0)) { fd_ctx->fd_status[i] = EC_FD_OPENING; - *need_open |= (1 << i); + need_open |= (1 << i); count++; } } @@ -76,10 +76,11 @@ ec_fd_ctx_need_open(fd_t *fd, xlator_t *this, uintptr_t *need_open) * then ignore fixing the fd as it has been * requested from heal operation. */ - if (count >= ec->fragments) - count = 0; + if (count >= ec->fragments) { + need_open = 0; + } - return count; + return need_open; } static gf_boolean_t @@ -96,9 +97,8 @@ ec_is_fd_fixable(fd_t *fd) } static void -ec_fix_open(ec_fop_data_t *fop) +ec_fix_open(ec_fop_data_t *fop, uintptr_t mask) { - int call_count = 0; uintptr_t need_open = 0; int ret = 0; loc_t loc = { @@ -109,9 +109,10 @@ ec_fix_open(ec_fop_data_t *fop) goto out; /* Evaluate how many remote fd's to be opened */ - call_count = ec_fd_ctx_need_open(fop->fd, fop->xl, &need_open); - if (!call_count) + need_open = ec_fd_ctx_need_open(fop->fd, fop->xl, mask); + if (need_open == 0) { goto out; + } loc.inode = inode_ref(fop->fd->inode); gf_uuid_copy(loc.gfid, fop->fd->inode->gfid); @@ -121,11 +122,13 @@ ec_fix_open(ec_fop_data_t *fop) } if (IA_IFDIR == fop->fd->inode->ia_type) { - ec_opendir(fop->frame, fop->xl, need_open, EC_MINIMUM_ONE, NULL, NULL, + ec_opendir(fop->frame, fop->xl, need_open, + EC_MINIMUM_ONE | EC_FOP_NO_PROPAGATE_ERROR, NULL, NULL, &fop->loc[0], fop->fd, NULL); } else { - ec_open(fop->frame, fop->xl, need_open, EC_MINIMUM_ONE, NULL, NULL, - &loc, fop->fd->flags, fop->fd, NULL); + ec_open(fop->frame, fop->xl, need_open, + EC_MINIMUM_ONE | EC_FOP_NO_PROPAGATE_ERROR, NULL, NULL, &loc, + fop->fd->flags, fop->fd, NULL); } out: @@ -495,12 +498,16 @@ ec_resume(ec_fop_data_t *fop, int32_t error) } void -ec_resume_parent(ec_fop_data_t *fop, int32_t error) +ec_resume_parent(ec_fop_data_t *fop) { ec_fop_data_t *parent; + int32_t error = 0; parent = fop->parent; if (parent != NULL) { + if ((fop->fop_flags & EC_FOP_NO_PROPAGATE_ERROR) == 0) { + error = fop->error; + } ec_trace("RESUME_PARENT", fop, "error=%u", error); fop->parent = NULL; ec_resume(parent, error); @@ -593,6 +600,8 @@ ec_internal_op(ec_fop_data_t *fop) return _gf_true; if (fop->id == GF_FOP_FXATTROP) return _gf_true; + if (fop->id == GF_FOP_OPEN) + return _gf_true; return _gf_false; } @@ -631,7 +640,7 @@ ec_msg_str(ec_fop_data_t *fop) return fop->errstr; } -int32_t +static int32_t ec_child_select(ec_fop_data_t *fop) { ec_t *ec = fop->xl->private; @@ -693,8 +702,6 @@ ec_child_select(ec_fop_data_t *fop) return 0; } - ec_sleep(fop); - return 1; } @@ -773,6 +780,8 @@ ec_dispatch_one(ec_fop_data_t *fop) ec_dispatch_start(fop); if (ec_child_select(fop)) { + ec_sleep(fop); + fop->expected = 1; fop->first = ec_select_first_by_read_policy(fop->xl->private, fop); @@ -807,6 +816,8 @@ ec_dispatch_inc(ec_fop_data_t *fop) ec_dispatch_start(fop); if (ec_child_select(fop)) { + ec_sleep(fop); + fop->expected = gf_bits_count(fop->remaining); fop->first = 0; @@ -820,6 +831,8 @@ ec_dispatch_all(ec_fop_data_t *fop) ec_dispatch_start(fop); if (ec_child_select(fop)) { + ec_sleep(fop); + fop->expected = gf_bits_count(fop->remaining); fop->first = 0; @@ -838,6 +851,8 @@ ec_dispatch_min(ec_fop_data_t *fop) ec_dispatch_start(fop); if (ec_child_select(fop)) { + ec_sleep(fop); + fop->expected = count = ec->fragments; fop->first = ec_select_first_by_read_policy(fop->xl->private, fop); idx = fop->first - 1; @@ -852,6 +867,23 @@ ec_dispatch_min(ec_fop_data_t *fop) } } +void +ec_succeed_all(ec_fop_data_t *fop) +{ + ec_dispatch_start(fop); + + if (ec_child_select(fop)) { + fop->expected = gf_bits_count(fop->remaining); + fop->first = 0; + + /* Simulate a successful execution on all bricks */ + ec_trace("SUCCEED", fop, ""); + + fop->good = fop->remaining; + fop->remaining = 0; + } +} + ec_lock_t * ec_lock_allocate(ec_fop_data_t *fop, loc_t *loc) { @@ -1825,7 +1857,8 @@ ec_lock_acquired(ec_lock_link_t *link) if (fop->use_fd && (link->update[EC_DATA_TXN] || link->update[EC_METADATA_TXN])) { - ec_fix_open(fop); + /* Try to reopen closed fd's only if lock has succeeded. */ + ec_fix_open(fop, lock->mask); } ec_lock_resume_shared(&list); |