| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
Fixes: bz#1734370
Change-Id: I29e338bac62104233a6f80212df8d0fb016affda
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
https://build.gluster.org/job/centos7-regression/7337/consoleFull
indicates the shd crashing for this .t. On looking at the core, I see
the crash is at the time of shd init and glusterfs context is null:
(gdb) bt
(gdb) p ctx
$2 = (glusterfs_ctx_t *) 0xf00000000
The .t is killing all gluster processes immediately after volume start,
so it looks like a race between shd coming up and it being killed.
Fix: Kill gluster processes only after they are up and running.
Fixes: bz#1740017
Change-Id: I7cf589201669bd9f535e968d147015dc99e9a4b6
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
tests/bugs/replicate/bug-1717819-metadata-split-brain-detection.t fails
intermittently in test cases #49 & #50, which compare the values of the
user set xattr values after enabling the heal. We are not waiting for
the heal to complete before comparing those values, which might lead
those tests to fail.
Fix:
Wait till the HEAL-TIMEOUT before comparing the xattr values.
Also cheking for the shd to come up and the bricks to connect to the shd
process in another case.
Change-Id: I0e245b328da9df23ce70c5300278fad1c1d9f7ff
Fixes: bz#1729847
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problems:
1. When checking for type and gfid mismatch, if the type or gfid
is unknown because of missing gfid handle and the gfid xattr
it will be reported as type or gfid mismatch and the heal will
not complete.
2. If the source selected during entry heal has null gfid the same
will be sent to afr_lookup_and_heal_gfid(). In this function when
we try to assign the gfid on the bricks where it does not exist,
we are considering the same gfid and try to assign that on those
bricks. This will fail in posix_gfid_set() since the gfid sent
is null.
Fix:
If the gfid sent to afr_lookup_and_heal_gfid() is null choose a
valid gfid before proceeding to assign the gfid on the bricks
where it is missing.
In afr_selfheal_detect_gfid_and_type_mismatch(), do not report
type/gfid mismatch if the type/gfid is unknown or not set.
Change-Id: Ia06552e4dc4a9f89cb7f5302833604bd21bbf7da
fixes: bz#1722507
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
We currently don't have a roll-back/undoing of post-ops if quorum is not met.
Though the FOP is still unwound with failure, the xattrs remain on the disk.
Due to these partial post-ops and partial heals (healing only when 2 bricks
are up), we can end up in metadata split-brain purely from the afr xattrs
point of view i.e each brick is blamed by atleast one of the others for
metadata. These scenarios are hit when there is frequent connect/disconnect
of the client/shd to the bricks.
Fix:
Pick a source based on the xattr values. If 2 bricks blame one, the blamed
one must be treated as sink. If there is no majority, all are sources. Once
we pick a source, self-heal will then do the heal instead of erroring out
due to split-brain.
This patch also adds restriction of all the bricks to be up to perform
metadata heal to avoid any metadata loss.
Removed the test case tests/bugs/replicate/bug-1468279-source-not-blaming-sinks.t
as it was doing metadata heal even when only 2 of 3 bricks were up.
Change-Id: I07a9d62f84ceda329dcab1f02a33aeed258dcb09
fixes: bz#1717819
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
| |
Change-Id: Iceefe22af754096c599dc570d4894d14fce4deae
Updates: bz#1193929
Signed-off-by: Xavier Hernandez <xhernandez@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When eager-lock lock acquisition fails because of say network failures, the
local is not being removed from owners_list, this leads to accumulation of
waiting frames and the application will hang because the waiting frames are
under the assumption that another transaction is in the process of acquiring
lock because owner-list is not empty. Handled this case as well in this patch.
Added asserts to make it easier to find these problems in future.
fixes bz#1696599
Change-Id: I3101393265e9827755725b1f2d94a93d8709e923
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Fixes afr_ta_read_txn() to handle inode refresh failures.
code-path.
- Fixes a double free issue of dict.
Note: This patch address post-merge review comments for commit
69532c141be160b3fea03c1579ae4ac13018dcdf
fixes: bz#1686398
Change-Id: Id5299b45b68569d47df6b73755918237a1592cb4
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
While checking a test case using EXPECT_WITHIN, the
argument is actually missing a '$' symbol to denote
the token as a variable in bash
Change-Id: I5b9150acdea000b29e94cfb01d975c77f5ece3e5
fixes: bz#1688116
Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
In an arbiter volume configuration SHD will not send any writes onto the arbiter
brick even if there is data pending marker for the arbiter brick. If we have a
arbiter setup on the geo-rep master and there are data pending markers for the files
on arbiter brick, SHD will not mark any data changelog during healing. While syncing
the data from master to slave, if the arbiter-brick is considered as ACTIVE, then
there is a chance that slave will miss out some data. If the arbiter brick is being
newly added or replaced there is a chance of slave missing all the data during sync.
Fix:
If there is data pending marker for the arbiter brick, send truncate on the arbiter
brick during heal, so that it will record truncate as the data transaction in changelog.
Change-Id: I3242ba6cea6da495c418ef860d9c3359c5459dec
fixes: bz#1686568
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
| |
Fixes: bz#1665358
Change-Id: Idbf88ec3ac683733b32c313377eeb72f2819bf0d
Signed-off-by: Amar Tumballi <amarts@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Automatic Splitbrain with size as policy must
not resolve splitbrains when the copies are of same size.
Determining if the sizes of copies are same and
returning -1 in that case.
updates: bz#1655052
Change-Id: I3d8e8b4d7962b070ed16c3ee02a1e5a926fd5eab
Signed-off-by: Iraj Jamali <ijamali@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In automatic Splitbrain resolution when favorite child policy
is set as size, split brain resolution must not work for
directories.
Currently, if a directory is in split brain with both copies
having same size, the source is selected arbitrarily
and healed.
fixes: bz#1655050
Change-Id: I5739498639c17c89874cc577362e543adab55f5d
Signed-off-by: Sheetal Pamecha <sheetal.pamecha08@gmail.com>
|
|
|
|
|
|
|
|
|
| |
With this changeset, default value for the AFR client side
heal volume option is set to "off"
fixes: bz#1663102
Change-Id: Ie4016932339c4896487e3e7cb5caca68739b7ba2
Signed-off-by: Sunil Kumar Acharya <sheggodu@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem: When trying to convert a plain distribute volume to replica-3
or arbiter type it is failing with ENOTCONN error as the lookup on
the root will fail as there is no quorum.
Fix: Allow lookup on root if it is coming from the ADD_REPLICA_MOUNT
which is used while adding bricks to a volume. It will try to set the
pending xattrs for the newly added bricks to allow the heal to happen
in the right direction and avoid data loss scenarios.
Note: This fix will solve the problem of type conversion only in the
case where the volume was mounted at least once. The conversion of
non mounted volumes will still fail since the dht selfheal tries to
set the directory layout will fail as they do that with the PID
GF_CLIENT_PID_NO_ROOT_SQUASH set in the frame->root.
Change-Id: Ic511939981dad118cc946754341318b164954b3b
fixes: bz#1655854
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Since rename/link fops on a file will not change any data in it, it should
not update the read_subvol values in the inode_ctx, which interprets the
data & metadata readable subvols for that file. The old read_subvol values
should be retained even after the rename/link operations.
Change-Id: I068044a426823a566f5bea8aa063cd689199d6dd
fixes: bz#1657783
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
If parent dir is in split-brain or has dirty xattrs set, and the file
has gfid missing on one of the bricks, then name heal won't assign the
gfid.
Fix:
Use the brick we select the gfid from as the 'source'.
Note: Problem was found while trying to debug a split-brain issue on
Cynthia Zhou's setup.
updates: bz#1637249
Change-Id: Id088d4f0fb017aa35122de426654194e581ed742
Reported-by: Cynthia Zhou <cynthia.zhou@nokia-sbell.com>
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Based on the proposal to remove few features as they are not
actively maintained [1], removing tier translator from the
build. Also make sure there are no regression tests involving
tiering feature are present.
[1] https://lists.gluster.org/pipermail/gluster-users/2018-July/034400.html
Change-Id: I2c177f711f9b54b7b24e1a13525ff3132bd9a9c5
updates: bz#1642807
Signed-off-by: Amar Tumballi <amarts@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
https://review.gluster.org/#/c/glusterfs/+/21427/ seems to be failing
this .t spuriously. On checking one of the failure logs, I see:
22:05:44 Launching heal operation to perform index self heal on volume patchy has been unsuccessful:
22:05:44 Self-heal daemon is not running. Check self-heal daemon log file.
22:05:44 not ok 20 , LINENUM:38
In glusterd log:
[2018-10-18 22:05:44.298832] E [MSGID: 106301] [glusterd-syncop.c:1352:gd_stage_op_phase] 0-management: Staging of operation 'Volume Heal' failed on localhost : Self-heal daemon is not running. Check self-heal daemon log file
But the tests which preceed this check whether via a statedump if the shd is
conected to the bricks, and they have succeeded and even started
healing. From glustershd.log:
[2018-10-18 22:05:40.975268] I [MSGID: 108026] [afr-self-heal-common.c:1732:afr_log_selfheal] 0-patchy-replicate-0: Completed data selfheal on 3b83d2dd-4cf2-4ea3-a33e-4275be40f440. sources=[0] 1 sinks=2
So the only reason I can see launching heal via cli failing is a race where
shd has been spawned but glusterd has not yet updated in-memory that it is up,
and hence failing the CLI.
Fix:
Check for shd up status before launching heal via CLI
Change-Id: Ic88abf14ad3d51c89cb438db601fae4df179e8f4
fixes: bz#1641344
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
In an arbiter volume, if there is a pending data heal of a file only on
arbiter brick, self-heal takes inodelks twice due to a code-bug but unlocks
it only once, leaving behind a stale lock on the brick. This causes
the next write to the file to hang.
Fix:
Fix the code-bug to take lock only once. This bug was introduced master
with commit eb472d82a083883335bc494b87ea175ac43471ff
Thanks to Pranith Kumar K <pkarampu@redhat.com> for finding the RCA.
fixes: bz#1637802
Change-Id: I15ad969e10a6a3c4bd255e2948b6be6dcddc61e1
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
When a directory has dirty xattrs due to failed post-ops or when
replace/reset brick is performed, AFR does a conservative merge as
expected, but heal-info reports it as split-brain because there are no
clear sources.
Fix:
Modify pending flag to contain information about pending heals and
split-brains. For directories, if spit-brain flag is not set,just show
them as needing heal and not being in split-brain.
Fixes: bz#1626994
Change-Id: I09ef821f6887c87d315ae99e6b1de05103cd9383
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
| |
- check that the shd is connected to brick before running statistics command
- remove sleep statements
- remove unneeded ($count-$value==0) test when it is known that both values will
be same
Fixes: bz#1625850
Change-Id: Ifcd4887f0238031e5bca803cd9bfdb75a6e6c01b
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
| |
Change-Id: Ia84cc24c8924e6d22d02ac15f611c10e26db99b4
Signed-off-by: Nigel Babu <nigelb@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
xlators/cluster/stripe/src/stripe-helpers.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
xlators/cluster/dht/src/tier.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
xlators/cluster/dht/src/dht-layout.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
xlators/cluster/dht/src/dht-helper.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
xlators/cluster/dht/src/dht-common.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
xlators/cluster/afr/src/afr.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
xlators/cluster/afr/src/afr-inode-read.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
tests/bugs/replicate/bug-1250170-fsync.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
tests/basic/gfapi/gfapi-async-calls-test.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
tests/basic/ec/ec-fast-fgetxattr.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
rpc/xdr/src/glusterfs3.h: Move to GF_MALLOC() instead of GF_CALLOC() when possible
rpc/rpc-transport/socket/src/socket.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
rpc/rpc-lib/src/rpc-clnt.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
extras/geo-rep/gsync-sync-gfid.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
cli/src/cli-xml-output.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
cli/src/cli-rpc-ops.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
cli/src/cli-cmd-volume.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
cli/src/cli-cmd-system.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
cli/src/cli-cmd-snapshot.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
cli/src/cli-cmd-peer.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
cli/src/cli-cmd-global.c: Move to GF_MALLOC() instead of GF_CALLOC() when possible
It doesn't make sense to calloc (allocate and clear) memory
when the code right away fills that memory with data.
It may be optimized by the compiler, or have a microscopic
performance improvement.
In some cases, also changed allocation size to be sizeof some
struct or type instead of a pointer - easier to read.
In some cases, removed redundant strlen() calls by saving the result
into a variable.
1. Only done for the straightforward cases. There's room for improvement.
2. Please review carefully, especially for string allocation, with the
terminating NULL string.
Only compile-tested!
updates: bz#1193929
Original-Author: Yaniv Kaul <ykaul@redhat.com>
Signed-off-by: Yaniv Kaul <ykaul@redhat.com>
Signed-off-by: Amar Tumballi <amarts@redhat.com>
Change-Id: I16274dca4078a1d06ae09a0daf027d734b631ac2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
In line #13 of the test case, it checks whether the file is present
on first 2 bricks or not. If it is not present on even one of the bricks
it will break the loop and checks for the dirty marking on the parent
on the 3rd brick and checks for file not present on the 1st and 2nd
bricks. The below scenario can happen in this case:
- File gets created on 1st and 3rd bricks
- In line #13 it sees file is not present on both 1st & 2nd bricks and
breaks the loop
- In line #51 test fails because the file will be present on the 1st brick
- In line #53 test will fail because the file creation was not failed on
quorum bricks and dirty marking will not be there on the parent on 3rd
brick
Fix:
Don't break from the loop if file is present on either brick 1 or brick 2.
Change-Id: I918068165e4b9124c1de86cfb373801b5b432bd9
fixes: bz#1612054
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
If the gfid link file inside .glusterfs is not present for a file,
the operations which are dependent on the gfid will fail,
complaining the link file does not exists inside .glusterfs.
Fix:
If the link file creation fails, fail the entry creation operation
and delete the original file.
Change-Id: Id767511de2da46b1f45aea45cb68b98d965ac96d
fixes: bz#1612037
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
The test case was checking for the entry pending marker reset
on the root after performing client side lookup at line #60-63.
But sometimes the entry heal was not getting completed immediately.
Fix:
Wait for the entry heal to complete before checking the changelog.
Change-Id: I42fde21b04a126ab044ce58373a996d72f125d96
fixes: bz#1614730
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
| |
See BZ for details.
Change-Id: I2cc2064f14d80271ebcc21747103ce4cee848cbf
fixes: bz#1615078
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
When we pass a command to be executed in EXPECT_WITHIN and we use ``
the value is passed by value, so if the first execution gives a result
that is different from the expected value, EXPECT_WITHIN test will
fail because the command will not be re-evaluated. Changed the
expression with `` to a function. Added sleep(3) in afr.c for
reconfigure to both RC and re-test after the change.
fixes bz#1614662
Change-Id: I3bc8a75b996729261aa48067f6ed8da9c6273b13
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
invalidation
Invalidations are triggered mainly by two codepaths - upcall and
write-behind unwinding a cached write with zeroed out stat. For the
case of upcall, following race can happen:
* stat s1 is fetched from brick
* invalidation is detected on brick
* invalidation is propagated to md-cache and cache is invalidated
* s1 updates md-cache with a stale state
For the case of write-behind, imagine following sequence of operations,
* A stat s1 was issued from application thread t1 when size of file
was s1
* stat s1 completes on brick stack, but yet to reach md-cache
* A write w1 from application thread t2 extends file to size s2 is
cached in write-behind and response is unwound with zeroed out stat
* md-cache while handling write-cbk, invalidates cache
* md-cache receives response for s1, updates cache with stale stat
with size of s1 overwriting invalidation state
Fix is to remember when s1 was incident on md-cache and update cache
with results of s1 only if the it was incident after invalidation of
cache.
This patch identified some bugs in regression tests which is tracked
in https://bugzilla.redhat.com/show_bug.cgi?id=1608158. As a stop gap
measure I am marking following tests as bad
basic/afr/split-brain-resolution.t
bugs/bug-1368312.t
bugs/replicate/bug-1238398-split-brain-resolution.t
bugs/replicate/bug-1417522-block-split-brain-resolution.t
bugs/replicate/bug-1438255-do-not-mark-self-accusing-xattrs.t
Change-Id: Ia4bb9dd36494944e2d91e9e71a79b5a3974a8c77
Signed-off-by: Raghavendra G <rgowdapp@redhat.com>
Updates: bz#1512691
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
If an entry creation transaction fails on quprum number of bricks
it might end up setting the pending changelogs on the file itself
on the brick where it got created. But the parent does not have
any entry pending marker set. This will lead to the entry not
getting healed by the self heal daemon automatically.
Fix:
For entry transactions mark dirty on the parent if it fails on
quorum number of bricks, so that the heal can do conservative
merge and entry gets healed by shd.
Change-Id: I56448932dd409b3ddb095e2ae32e037b6157a607
fixes: bz#1586020
Signed-off-by: karthik-us <ksubrahm@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
1) snprintf into linkname_expected should happen with PATH_MAX
2) comparison should happen with linkname_actual with complete
string linkname_expected
fixes bz#1595190
Change-Id: Ic3b3c362dc6c69c046b9a13e031989be47ecff14
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
| |
commit 20fa80057eb430fd72b4fa31b9b65598b8ec1265 introduced a regression
wherein if a file is present in only 1 brick of replica *and* doesn't
have a gfid associated with it, it doesn't get healed upon the next
lookup from the client. Fix it.
Change-Id: I7d1111dcb45b1b8b8340a7d02558f05df70aa599
fixes: bz#1591193
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
In the .t, when the only good brick was brought down, writes on the fd were
still succeeding on the bad bricks. The inflight split-brain check was
marking the write as failure but since the write succeeded on all the
bad bricks, afr_txn_nothing_failed() was set to true and we were
unwinding writev with success to DHT and then catching the failure in
post-op in the background.
Fix:
Don't wind the FOP phase if the write_subvol (which is populated with readable
subvols obtained in pre-op cbk) does not have at least 1 good brick which was up
when the transaction started.
Note: This fix is not related to brick muliplexing. I ran the .t
10 times with this fix and brick-mux enabled without any failures.
Change-Id: I915c9c366aa32cd342b1565827ca2d83cb02ae85
updates: bz#1577672
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
1. If pre-op fails on all bricks,set lock->release to true in
afr_handle_lock_acquire_failure so that the GF_ASSERT in afr_unlock() does not
crash.
2. Added a missing 'return' after handling pre-op failure in
afr_transaction_perform_fop(), fixing a use-after-free issue.
Change-Id: If0627a9124cb5d6405037cab3f17f8325eed2d83
fixes: bz#1561129
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
1) Afr's eager-lock only works for data transactions.
2) When there are conflicting writes, write with conflicting region initiates
unlock of eager-lock leading to extra pre-ops and post-ops on the file. When
eager-lock goes off, it leads to extra fsyncs for random-write workload in afr.
Solution (that is modeled after EC):
In EC, when there is a conflicting write, it waits for the current write to
complete before it winds the conflicted write. This leads to better utilization
of network and disk, because we will not be doing extra xattrops and FSYNCs and
inodelk/unlock. Moved fd based counters to inode based counters.
I tried to model the solution based on EC's locking, but it is not similar to
AFR because we had to keep backward compatibility.
Lifecycle of lock:
==================
First transaction is added to inode->owners list and an inodelk will be sent on
the wire. All the next transactions will be put in inode->waiters list until
the first transaction completes inodelk and [f]xattrop completely. Once
[f]xattrop also completes, all the requests in the inode->waiters list are
checked if it conflict with any of the existing locks which are in
inode->owners list and if not are added to inode->owners list and resumed with
doing transaction. When these transactions complete fop phase they will be
moved to inode->post_op list and resume the transactions that were paused
because of conflicts. Post-op and unlock will not be issued on the wire until
that is the last transaction on that inode. Last transaction when it has to
perform post-op can choose to sleep for deyed-post-op-secs value. During that
time if any other transaction comes, it will wake up the sleeping transaction
and takes over the ownership of the lock and the cycle continues. If the
dealyed-post-op-secs expire, then the timer thread will wakeup the sleeping
transaction and it will set lock->release to true and starts doing post-op and
then unlock. During this time if any other transactions come, they will be put
in inode->frozen list. Once the previous unlock comes it will move the frozen
list to waiters list and moves the first element from this waiters-list to
owners-list and attempts the lock and the cycle continues. This is the general
idea. There is logic at the time of dealying and at the time of new
transaction or in flush fop to wakeup existing sleeping transactions or
choosing whether to delay a transaction etc, which is subjected to change based
on future enhancements etc.
Fixes: #418
BUG: 1549606
Change-Id: I88b570bbcf332a27c82d2767dfa82472f60055dc
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
We currently don't have a roll-back/undoing of post-ops if quorum is not
met. Though the FOP is still unwound with failure, the xattrs remain on
the disk. Due to these partial post-ops and partial heals (healing only when
2 bricks are up), we can end up in split-brain purely from the afr
xattrs point of view i.e each brick is blamed by atleast one of the
others. These scenarios are hit when there is frequent
connect/disconnect of the client/shd to the bricks while I/O or heal
are in progress.
Fix:
Instead of undoing the post-op, pick a source based on the xattr values.
If 2 bricks blame one, the blamed one must be treated as sink.
If there is no majority, all are sources. Once we pick a source,
self-heal will then do the heal instead of erroring out due to
split-brain.
Change-Id: I3d0224b883eb0945785ade0e9697a1c828aec0ae
BUG: 1539358
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
afr relies on pending changelog xattrs to identify source and sinks and the
setting of these xattrs happen in post-op. So if post-op fails, we need to
unwind the write txn with a failure.
Change-Id: I0f019ac03890108324ee7672883d774918b20be1
BUG: 1506140
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
so that glusterd is also aware that shd is up and running.
While not reproducible locally, on the jenkins slaves, 'gluster vol heal patchy'
fails with "Self-heal daemon is not running. Check self-heal daemon log file.",
while infact the afr_child_up_status_in_shd() checks before that passed. In the
shd log also, I see the shd being up and connected to at least one brick before
the heal is launched.
Change-Id: Id3801fa4ab56a70b1f0bd6a7e240f69bea74a5fc
BUG: 1515163
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
| |
This reverts commit 56e5fdae74845dfec0ff7ad0c8fee77695d36ad5.
Change-Id: Ia62cee5440bbe8e23f5da9cff692d792091d544a
Signed-off-by: Milind Changire <mchangir@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Patch attempts to use the epoll infra for handling SSL connections
as well instead of the socket_poller() thread func.
This essentially makes priv->own_thread flag redundant.
SSL_connect()/SSL_accept() is now non-blocking which has done away
with the localised poll() in ssl_do(). So, ssl_do() has been updated
appropriately.
own_thread and coincidently socket_poller() thread for SSL processing
is now deprecated.
Added a timeout to test whether seal-heal daemon is up and running
as per Ravi's suggestion.
Change-Id: If2b5d7b4fd19e321cb289e08d49a718d2161aafe
Signed-off-by: Milind Changire <mchangir@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
In an arbiter volume, lookup was being served from one of the sink
bricks (source brick was down). shard uses the iatt values from lookup cbk
to calculate the size and block count, which in this case were incorrect
values. shard_local_t->last_block was thus initialised to -1, resulting
in an infinite while loop in shard_common_resolve_shards().
Fix:
Use client quorum logic to allow or fail the lookups from afr if there
are no readable subvolumes. So in replica-3 or arbiter vols, if there is
no good copy or if quorum is not met, fail lookup with ENOTCONN.
With this fix, we are also removing support for quorum-reads xlator
option. So if quorum is not met, neither read nor write txns are allowed
and we fail the fop with ENOTCONN.
Change-Id: Ic65c00c24f77ece007328b421494eee62a505fa0
BUG: 1467250
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The restriction of using fds opened by the same Pid means fds cannot
be shared across threads of multithreaded application. Note that fops
from kernel have different Pid for different threads. Imagine
following sequence of operations:
* Turn off performance.open-behind
* Thread t1 opens an fd - fd1 - on file "file". Let's assume nodeid of
"file" is "nodeid-file".
* Thread t2 does RENAME ("newfile", "file"). Let's assume nodeid of
"newfile" as "nodeid-newfile".
* t2 proceeds to do fstat (fd1)
The above set of operations can sometimes result in ESTALE/ENOENT
errors. RENAME overwrites "file" with "newfile" changing its nodeid
from "nodeid-file" to "nodeid-newfile" and post RENAME, "nodeid-file" is
removed from the backend. If fstat carries nodeid-file as argument,
which can happen if lookup has not refreshed the nodeid of "file" and
since t2 doesn't have an fd opened, fuse_getattr_resume uses STAT
which will fail as "nodeid-file" no longer exists.
Since the above set of operations and sharing of fds across
multiple threads are valid, this is a bug.
The fix is to use any fd opened on the inode. In this specific example
fuse_getattr_resume will find fd1 and winds down the call as fstat
(fd1) which won't fail.
Cross-checked with "Miklos Szeredi" <mszeredi.at.redhat.dot.com> for
any security issues with this solution and he approves the solution.
Thanks to "Miklos Szeredi" <mszeredi.at.redhat.dot.com> for all the
pointers and discussions.
Change-Id: I88dd29b3607cd2594eee9d72a1637b5346c8d49c
BUG: 1510401
Signed-off-by: Raghavendra G <rgowdapp@redhat.com>
|
|
|
|
|
|
|
|
| |
Update .t tier tests to use the new tier CLI.
Change-Id: I0e7f1769071108d8266fc86378c4466bcaf96e7d
BUG: 1505253
Signed-off-by: N Balachandran <nbalacha@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
Commit ff075a3d6f9b142911d25c27fd209838782bfff0 disabled loading
client-io-threads for replicate volumes (it was set to on by default in
commit e068c1997314046658dd502e9118dab32decf879) due to performance
issues but in doing so, inadvertently failed to load the xlator even if
the user explicitly enabled the option using the volume set command.
This was despite returning returning sucess for the volume set.
Fix:
Modify the check in perfxl_option_handler() and add checks in volume
create/add-brick/remove-brick code paths, tying it all to
GD_OP_VERSION_3_12_2.
Change-Id: Ib612973a999a7da818cc926f5c2601b1f0794fcf
BUG: 1498570
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problem:
If a brick crashes after an entry (file or dir) is created but before
gfid is assigned, the good bricks will have pending entry heal xattrs
but the heal won't complete because afr_selfheal_recreate_entry() tries
to create the entry again and it fails with EEXIST.
Fix:
We could have fixed posx_mknod/mkdir etc to assign the gfid if the file
already exists but the right thing to do seems to be to trigger a lookup
on the bad brick and let it heal the gfid instead of winding an
mknod/mkdir in the first place.
Change-Id: I82f76665a7541f1893ef8d847b78af6466aff1ff
BUG: 1493415
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
|
|
|
|
|
|
|
|
|
|
| |
With this change, enabling choose-local (which means its state makes
transition from "off" to "on") will be effective after the first
gfid-lookup on "/" since volume-set was executed.
Change-Id: Ibab292ba705d993b475cd0303fb3318211fb2500
BUG: 1480525
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Problems:
As described in BZ 1491670, renaming hardlinks can result in data/mdata
split-brain of the DHT link-to files (T files) without any mismatch of
data and metadata.
As described in BZ 1486063, for a zero-byte file with only dirty bits
set, arbiter brick will likely be chosen as the source brick.
Fix:
For zero byte files in split-brain, pick first brick as
a) data source if file size is zero on all bricks.
b) metadata source if metadata is the same on all bricks
In arbiter case, if file size is zero on all bricks and there are no
pending afr xattrs, pick 1st brick as data source.
Change-Id: I0270a9a2f97c3b21087e280bb890159b43975e04
BUG: 1491670
Signed-off-by: Ravishankar N <ravishankar@redhat.com>
Reported-by: Rahul Hinduja <rhinduja@redhat.com>
Reported-by: Mabi <mabi@protonmail.ch>
|
|
|
|
|
|
|
|
|
|
|
| |
Change-Id: I55fa87e07136cff10b0d725ee24dd3151016e64e
BUG: 1489823
Signed-off-by: Krutika Dhananjay <kdhananj@redhat.com>
Reviewed-on: https://review.gluster.org/18243
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
Smoke: Gluster Build System <jenkins@build.gluster.org>
Tested-by: Sunil Kumar Acharya <sheggodu@redhat.com>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
glusterd shouldn't concern itself with creating directories specific to
certain xlators.
The index xlator will now proceed creating './glusterfs/indices' dir
only if the parent '.glusterfs' directory exists, which still fixes the
original problem reported i.e 'volume start force' command shouldn't
create brick path if it doesn't exist (BUG 1457202)
This reverts most of the changes done by the commit
b58a15948fb3fc37b6c0b70171482f50ed957f42
Change-Id: I7fc52ad64dce220e336c218fb4d85933ca2e61c0
Signed-off-by: Prashanth Pai <ppai@redhat.com>
Reviewed-on: https://review.gluster.org/18003
Smoke: Gluster Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.org>
Reviewed-by: Jeff Darcy <jeff@pl.atyp.us>
Reviewed-by: Pranith Kumar Karampuri <pkarampu@redhat.com>
|