summaryrefslogtreecommitdiff
path: root/0439-cluster-afr-Delay-post-op-for-fsync.patch
blob: dc1593bea61986f2d2e2b31a51747bb7747ea4f8 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
From 3ed98fc9dcb39223032e343fd5b0ad17fa3cae14 Mon Sep 17 00:00:00 2001
From: Pranith Kumar K <pkarampu@redhat.com>
Date: Fri, 29 May 2020 14:24:53 +0530
Subject: [PATCH 439/449] cluster/afr: Delay post-op for fsync

Problem:
AFR doesn't delay post-op for fsync fop. For fsync heavy workloads
this leads to un-necessary fxattrop/finodelk for every fsync leading
to bad performance.

Fix:
Have delayed post-op for fsync. Add special flag in xdata to indicate
that afr shouldn't delay post-op in cases where either the
process will terminate or graph-switch would happen. Otherwise it leads
to un-necessary heals when the graph-switch/process-termination
happens before delayed-post-op completes.

> Upstream-patch: https://review.gluster.org/c/glusterfs/+/24473
> Fixes: #1253

BUG: 1838479
Change-Id: I531940d13269a111c49e0510d49514dc169f4577
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/202676
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Sunil Kumar Heggodu Gopala Acharya <sheggodu@redhat.com>
---
 api/src/glfs-resolve.c                         |  14 ++-
 tests/basic/afr/durability-off.t               |   2 +
 tests/basic/gfapi/gfapi-graph-switch-open-fd.t |  44 +++++++++
 tests/basic/gfapi/gfapi-keep-writing.c         | 129 +++++++++++++++++++++++++
 xlators/cluster/afr/src/afr-inode-write.c      |  11 ++-
 xlators/cluster/afr/src/afr-transaction.c      |   9 +-
 xlators/cluster/afr/src/afr.h                  |   2 +-
 xlators/cluster/dht/src/dht-rebalance.c        |  15 ++-
 xlators/mount/fuse/src/fuse-bridge.c           |  23 ++++-
 9 files changed, 239 insertions(+), 10 deletions(-)
 create mode 100644 tests/basic/gfapi/gfapi-graph-switch-open-fd.t
 create mode 100644 tests/basic/gfapi/gfapi-keep-writing.c

diff --git a/api/src/glfs-resolve.c b/api/src/glfs-resolve.c
index a79f490..062b7dc 100644
--- a/api/src/glfs-resolve.c
+++ b/api/src/glfs-resolve.c
@@ -722,6 +722,7 @@ glfs_migrate_fd_safe(struct glfs *fs, xlator_t *newsubvol, fd_t *oldfd)
         0,
     };
     char uuid1[64];
+    dict_t *xdata = NULL;
 
     oldinode = oldfd->inode;
     oldsubvol = oldinode->table->xl;
@@ -730,7 +731,15 @@ glfs_migrate_fd_safe(struct glfs *fs, xlator_t *newsubvol, fd_t *oldfd)
         return fd_ref(oldfd);
 
     if (!oldsubvol->switched) {
-        ret = syncop_fsync(oldsubvol, oldfd, 0, NULL, NULL, NULL, NULL);
+        xdata = dict_new();
+        if (!xdata || dict_set_int8(xdata, "last-fsync", 1)) {
+            gf_msg(fs->volname, GF_LOG_WARNING, ENOMEM, API_MSG_FSYNC_FAILED,
+                   "last-fsync set failed on %s graph %s (%d)",
+                   uuid_utoa_r(oldfd->inode->gfid, uuid1),
+                   graphid_str(oldsubvol), oldsubvol->graph->id);
+        }
+
+        ret = syncop_fsync(oldsubvol, oldfd, 0, NULL, NULL, xdata, NULL);
         DECODE_SYNCOP_ERR(ret);
         if (ret) {
             gf_msg(fs->volname, GF_LOG_WARNING, errno, API_MSG_FSYNC_FAILED,
@@ -809,6 +818,9 @@ out:
         newfd = NULL;
     }
 
+    if (xdata)
+        dict_unref(xdata);
+
     return newfd;
 }
 
diff --git a/tests/basic/afr/durability-off.t b/tests/basic/afr/durability-off.t
index 155ffa0..6e0f18b 100644
--- a/tests/basic/afr/durability-off.t
+++ b/tests/basic/afr/durability-off.t
@@ -26,6 +26,8 @@ TEST $CLI volume heal $V0
 EXPECT_WITHIN $HEAL_TIMEOUT "0" get_pending_heal_count $V0
 EXPECT "^0$" echo $($CLI volume profile $V0 info | grep -w FSYNC | wc -l)
 
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 0
+EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" afr_child_up_status $V0 1
 #Test that fsyncs happen when durability is on
 TEST $CLI volume set $V0 cluster.ensure-durability on
 TEST $CLI volume set $V0 performance.strict-write-ordering on
diff --git a/tests/basic/gfapi/gfapi-graph-switch-open-fd.t b/tests/basic/gfapi/gfapi-graph-switch-open-fd.t
new file mode 100644
index 0000000..2e666be
--- /dev/null
+++ b/tests/basic/gfapi/gfapi-graph-switch-open-fd.t
@@ -0,0 +1,44 @@
+#!/bin/bash
+
+. $(dirname $0)/../../include.rc
+. $(dirname $0)/../../volume.rc
+
+cleanup;
+
+TEST glusterd
+
+TEST $CLI volume create $V0 replica 3 ${H0}:$B0/brick{0..2};
+EXPECT 'Created' volinfo_field $V0 'Status';
+
+TEST $CLI volume start $V0;
+EXPECT 'Started' volinfo_field $V0 'Status';
+
+TEST $GFS --volfile-id=$V0 --volfile-server=$H0 $M0;
+TEST touch $M0/sync
+logdir=`gluster --print-logdir`
+
+TEST build_tester $(dirname $0)/gfapi-keep-writing.c -lgfapi
+
+
+#Launch a program to keep doing writes on an fd
+./$(dirname $0)/gfapi-keep-writing ${H0} $V0 $logdir/gfapi-async-calls-test.log sync &
+p=$!
+sleep 1 #Let some writes go through
+#Check if graph switch will lead to any pending markers for ever
+TEST $CLI volume set $V0 performance.quick-read off
+TEST $CLI volume set $V0 performance.io-cache off
+TEST $CLI volume set $V0 performance.stat-prefetch off
+TEST $CLI volume set $V0 performance.read-ahead off
+
+
+TEST rm -f $M0/sync #Make sure the glfd is closed
+TEST wait #Wait for background process to die
+#Goal is to check if there is permanent FOOL changelog
+sleep 5
+EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/brick0/glfs_test.txt trusted.afr.dirty
+EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/brick1/glfs_test.txt trusted.afr.dirty
+EXPECT "0x000000000000000000000000" afr_get_changelog_xattr $B0/brick2/glfs_test.txt trusted.afr.dirty
+
+cleanup_tester $(dirname $0)/gfapi-async-calls-test
+
+cleanup;
diff --git a/tests/basic/gfapi/gfapi-keep-writing.c b/tests/basic/gfapi/gfapi-keep-writing.c
new file mode 100644
index 0000000..91b59ce
--- /dev/null
+++ b/tests/basic/gfapi/gfapi-keep-writing.c
@@ -0,0 +1,129 @@
+#include <fcntl.h>
+#include <unistd.h>
+#include <time.h>
+#include <limits.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <errno.h>
+#include <glusterfs/api/glfs.h>
+#include <glusterfs/api/glfs-handles.h>
+
+#define LOG_ERR(msg)                                                           \
+    do {                                                                       \
+        fprintf(stderr, "%s : Error (%s)\n", msg, strerror(errno));            \
+    } while (0)
+
+glfs_t *
+init_glfs(const char *hostname, const char *volname, const char *logfile)
+{
+    int ret = -1;
+    glfs_t *fs = NULL;
+
+    fs = glfs_new(volname);
+    if (!fs) {
+        LOG_ERR("glfs_new failed");
+        return NULL;
+    }
+
+    ret = glfs_set_volfile_server(fs, "tcp", hostname, 24007);
+    if (ret < 0) {
+        LOG_ERR("glfs_set_volfile_server failed");
+        goto out;
+    }
+
+    ret = glfs_set_logging(fs, logfile, 7);
+    if (ret < 0) {
+        LOG_ERR("glfs_set_logging failed");
+        goto out;
+    }
+
+    ret = glfs_init(fs);
+    if (ret < 0) {
+        LOG_ERR("glfs_init failed");
+        goto out;
+    }
+
+    ret = 0;
+out:
+    if (ret) {
+        glfs_fini(fs);
+        fs = NULL;
+    }
+
+    return fs;
+}
+
+int
+glfs_test_function(const char *hostname, const char *volname,
+                   const char *logfile, const char *syncfile)
+{
+    int ret = -1;
+    int flags = O_CREAT | O_RDWR;
+    glfs_t *fs = NULL;
+    glfs_fd_t *glfd = NULL;
+    const char *buff = "This is from my prog\n";
+    const char *filename = "glfs_test.txt";
+    struct stat buf = {0};
+
+    fs = init_glfs(hostname, volname, logfile);
+    if (fs == NULL) {
+        LOG_ERR("init_glfs failed");
+        return -1;
+    }
+
+    glfd = glfs_creat(fs, filename, flags, 0644);
+    if (glfd == NULL) {
+        LOG_ERR("glfs_creat failed");
+        goto out;
+    }
+
+    while (glfs_stat(fs, syncfile, &buf) == 0) {
+        ret = glfs_write(glfd, buff, strlen(buff), flags);
+        if (ret < 0) {
+            LOG_ERR("glfs_write failed");
+            goto out;
+        }
+    }
+
+    ret = glfs_close(glfd);
+    if (ret < 0) {
+        LOG_ERR("glfs_write failed");
+        goto out;
+    }
+
+out:
+    ret = glfs_fini(fs);
+    if (ret) {
+        LOG_ERR("glfs_fini failed");
+    }
+
+    return ret;
+}
+
+int
+main(int argc, char *argv[])
+{
+    int ret = 0;
+    char *hostname = NULL;
+    char *volname = NULL;
+    char *logfile = NULL;
+    char *syncfile = NULL;
+
+    if (argc != 5) {
+        fprintf(stderr, "Invalid argument\n");
+        exit(1);
+    }
+
+    hostname = argv[1];
+    volname = argv[2];
+    logfile = argv[3];
+    syncfile = argv[4];
+
+    ret = glfs_test_function(hostname, volname, logfile, syncfile);
+    if (ret) {
+        LOG_ERR("glfs_test_function failed");
+    }
+
+    return ret;
+}
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 7fcc9d4..df82b6e 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -2492,6 +2492,7 @@ afr_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
     call_frame_t *transaction_frame = NULL;
     int ret = -1;
     int32_t op_errno = ENOMEM;
+    int8_t last_fsync = 0;
 
     transaction_frame = copy_frame(frame);
     if (!transaction_frame)
@@ -2501,10 +2502,16 @@ afr_fsync(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
     if (!local)
         goto out;
 
-    if (xdata)
+    if (xdata) {
         local->xdata_req = dict_copy_with_ref(xdata, NULL);
-    else
+        if (dict_get_int8(xdata, "last-fsync", &last_fsync) == 0) {
+            if (last_fsync) {
+                local->transaction.disable_delayed_post_op = _gf_true;
+            }
+        }
+    } else {
         local->xdata_req = dict_new();
+    }
 
     if (!local->xdata_req)
         goto out;
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index 8e65ae2..ffd0ab8 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -2385,8 +2385,13 @@ afr_is_delayed_changelog_post_op_needed(call_frame_t *frame, xlator_t *this,
         goto out;
     }
 
-    if ((local->op != GF_FOP_WRITE) && (local->op != GF_FOP_FXATTROP)) {
-        /*Only allow writes but shard does [f]xattrops on writes, so
+    if (local->transaction.disable_delayed_post_op) {
+        goto out;
+    }
+
+    if ((local->op != GF_FOP_WRITE) && (local->op != GF_FOP_FXATTROP) &&
+        (local->op != GF_FOP_FSYNC)) {
+        /*Only allow writes/fsyncs but shard does [f]xattrops on writes, so
          * they are fine too*/
         goto out;
     }
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 18f1a6a..ff96246 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -854,7 +854,7 @@ typedef struct _afr_local {
 
         int (*unwind)(call_frame_t *frame, xlator_t *this);
 
-        /* post-op hook */
+        gf_boolean_t disable_delayed_post_op;
     } transaction;
 
     syncbarrier_t barrier;
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
index d0c21b4..e9974cd 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -1550,6 +1550,7 @@ dht_migrate_file(xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
     xlator_t *old_target = NULL;
     xlator_t *hashed_subvol = NULL;
     fd_t *linkto_fd = NULL;
+    dict_t *xdata = NULL;
 
     if (from == to) {
         gf_msg_debug(this->name, 0,
@@ -1868,7 +1869,15 @@ dht_migrate_file(xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
 
     /* TODO: Sync the locks */
 
-    ret = syncop_fsync(to, dst_fd, 0, NULL, NULL, NULL, NULL);
+    xdata = dict_new();
+    if (!xdata || dict_set_int8(xdata, "last-fsync", 1)) {
+        gf_log(this->name, GF_LOG_ERROR,
+               "%s: failed to set last-fsync flag on "
+               "%s (%s)",
+               loc->path, to->name, strerror(ENOMEM));
+    }
+
+    ret = syncop_fsync(to, dst_fd, 0, NULL, NULL, xdata, NULL);
     if (ret) {
         gf_log(this->name, GF_LOG_WARNING, "%s: failed to fsync on %s (%s)",
                loc->path, to->name, strerror(-ret));
@@ -2342,11 +2351,15 @@ out:
 
     if (dst_fd)
         syncop_close(dst_fd);
+
     if (src_fd)
         syncop_close(src_fd);
     if (linkto_fd)
         syncop_close(linkto_fd);
 
+    if (xdata)
+        dict_unref(xdata);
+
     loc_wipe(&tmp_loc);
     loc_wipe(&parent_loc);
 
diff --git a/xlators/mount/fuse/src/fuse-bridge.c b/xlators/mount/fuse/src/fuse-bridge.c
index fdeec49..4264fad 100644
--- a/xlators/mount/fuse/src/fuse-bridge.c
+++ b/xlators/mount/fuse/src/fuse-bridge.c
@@ -5559,6 +5559,7 @@ fuse_migrate_fd(xlator_t *this, fd_t *basefd, xlator_t *old_subvol,
     char create_in_progress = 0;
     fuse_fd_ctx_t *basefd_ctx = NULL;
     fd_t *oldfd = NULL;
+    dict_t *xdata = NULL;
 
     basefd_ctx = fuse_fd_ctx_get(this, basefd);
     GF_VALIDATE_OR_GOTO("glusterfs-fuse", basefd_ctx, out);
@@ -5595,10 +5596,23 @@ fuse_migrate_fd(xlator_t *this, fd_t *basefd, xlator_t *old_subvol,
     }
 
     if (oldfd->inode->table->xl == old_subvol) {
-        if (IA_ISDIR(oldfd->inode->ia_type))
+        if (IA_ISDIR(oldfd->inode->ia_type)) {
             ret = syncop_fsyncdir(old_subvol, oldfd, 0, NULL, NULL);
-        else
-            ret = syncop_fsync(old_subvol, oldfd, 0, NULL, NULL, NULL, NULL);
+        } else {
+            xdata = dict_new();
+            if (!xdata || dict_set_int8(xdata, "last-fsync", 1)) {
+                gf_log("glusterfs-fuse", GF_LOG_WARNING,
+                       "last-fsync set failed (%s) on fd (%p)"
+                       "(basefd:%p basefd-inode.gfid:%s) "
+                       "(old-subvolume:%s-%d new-subvolume:%s-%d)",
+                       strerror(ENOMEM), oldfd, basefd,
+                       uuid_utoa(basefd->inode->gfid), old_subvol->name,
+                       old_subvol->graph->id, new_subvol->name,
+                       new_subvol->graph->id);
+            }
+
+            ret = syncop_fsync(old_subvol, oldfd, 0, NULL, NULL, xdata, NULL);
+        }
 
         if (ret < 0) {
             gf_log("glusterfs-fuse", GF_LOG_WARNING,
@@ -5653,6 +5667,9 @@ out:
 
     fd_unref(oldfd);
 
+    if (xdata)
+        dict_unref(xdata);
+
     return ret;
 }
 
-- 
1.8.3.1