diff options
| -rw-r--r-- | Makefile.am | 2 | ||||
| -rw-r--r-- | cli/src/cli-cmd-volume.c | 31 | ||||
| -rw-r--r-- | configure.ac | 2 | ||||
| -rw-r--r-- | glusterfs.spec.in | 1 | ||||
| -rw-r--r-- | heal/Makefile.am | 3 | ||||
| -rw-r--r-- | heal/src/Makefile.am | 28 | ||||
| -rw-r--r-- | heal/src/glfs-heal.c | 488 |
7 files changed, 550 insertions, 5 deletions
diff --git a/Makefile.am b/Makefile.am index 598ebb4103c..19f21fca8a2 100644 --- a/Makefile.am +++ b/Makefile.am @@ -7,7 +7,7 @@ EXTRA_DIST = autogen.sh \ $(shell find $(top_srcdir)/tests -type f -print) SUBDIRS = argp-standalone libglusterfs rpc api xlators glusterfsd \ - $(FUSERMOUNT_SUBDIR) doc extras cli @SYNCDAEMON_SUBDIR@ + $(FUSERMOUNT_SUBDIR) doc extras cli heal @SYNCDAEMON_SUBDIR@ pkgconfigdir = @pkgconfigdir@ pkgconfig_DATA = glusterfs-api.pc libgfchangelog.pc diff --git a/cli/src/cli-cmd-volume.c b/cli/src/cli-cmd-volume.c index 22bf66b4fb5..80b559dc69f 100644 --- a/cli/src/cli-cmd-volume.c +++ b/cli/src/cli-cmd-volume.c @@ -2033,6 +2033,10 @@ cli_cmd_volume_heal_cbk (struct cli_state *state, struct cli_cmd_word *word, dict_t *options = NULL; xlator_t *this = NULL; cli_local_t *local = NULL; + int heal_op = 0; + runner_t runner = {0}; + char buff[PATH_MAX] = {0}; + char *out = NULL; this = THIS; frame = create_frame (this, this->ctx->pool); @@ -2051,13 +2055,32 @@ cli_cmd_volume_heal_cbk (struct cli_state *state, struct cli_cmd_word *word, parse_error = 1; goto out; } + ret = dict_get_int32 (options, "heal-op", &heal_op); + if (ret < 0) + goto out; - proc = &cli_rpc_prog->proctable[GLUSTER_CLI_HEAL_VOLUME]; + if (heal_op == GF_AFR_OP_INDEX_SUMMARY) { + runinit (&runner); + runner_add_args (&runner, SBIN_DIR"/glfsheal", words[2], NULL); + runner_redir (&runner, STDOUT_FILENO, RUN_PIPE); + ret = runner_start (&runner); + if (ret == -1) + goto out; + while ((out = fgets(buff, sizeof(buff), + runner_chio (&runner, STDOUT_FILENO)))) { + printf ("%s", out); + } - CLI_LOCAL_INIT (local, words, frame, options); + ret = runner_end (&runner); + ret = WEXITSTATUS (ret); + } else { + proc = &cli_rpc_prog->proctable[GLUSTER_CLI_HEAL_VOLUME]; - if (proc->fn) { - ret = proc->fn (frame, THIS, options); + CLI_LOCAL_INIT (local, words, frame, options); + + if (proc->fn) { + ret = proc->fn (frame, THIS, options); + } } out: diff --git a/configure.ac b/configure.ac index f1a9e6da267..37998e60d77 100644 --- a/configure.ac +++ b/configure.ac @@ -179,6 +179,8 @@ AC_CONFIG_FILES([Makefile geo-replication/Makefile geo-replication/src/Makefile geo-replication/syncdaemon/Makefile + heal/Makefile + heal/src/Makefile glusterfs.spec]) AC_CANONICAL_HOST diff --git a/glusterfs.spec.in b/glusterfs.spec.in index 65e39b677e6..6b65957fb6a 100644 --- a/glusterfs.spec.in +++ b/glusterfs.spec.in @@ -789,6 +789,7 @@ fi %endif # binaries %{_sbindir}/glusterd +%{_sbindir}/glfsheal %{_libdir}/glusterfs/%{version}%{?prereltag}/xlator/storage* %{_libdir}/glusterfs/%{version}%{?prereltag}/xlator/features/posix* %{_libdir}/glusterfs/%{version}%{?prereltag}/xlator/protocol/server* diff --git a/heal/Makefile.am b/heal/Makefile.am new file mode 100644 index 00000000000..a985f42a877 --- /dev/null +++ b/heal/Makefile.am @@ -0,0 +1,3 @@ +SUBDIRS = src + +CLEANFILES = diff --git a/heal/src/Makefile.am b/heal/src/Makefile.am new file mode 100644 index 00000000000..2ea41b969f6 --- /dev/null +++ b/heal/src/Makefile.am @@ -0,0 +1,28 @@ +sbin_PROGRAMS = glfsheal + +glfsheal_SOURCES = glfs-heal.c + +glfsheal_LDADD = $(top_builddir)/libglusterfs/src/libglusterfs.la $(GF_LDADD)\ + $(RLLIBS) $(top_builddir)/rpc/xdr/src/libgfxdr.la \ + $(top_builddir)/rpc/rpc-lib/src/libgfrpc.la \ + $(top_builddir)/api/src/libgfapi.la\ + $(GF_GLUSTERFS_LIBS) $(XML_LIBS) + +glfsheal_LDFLAGS = $(GF_LDFLAGS) + +AM_CPPFLAGS = $(GF_CPPFLAGS) \ + -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/rpc/rpc-lib/src\ + -I$(top_srcdir)/rpc/xdr/src\ + -I$(top_srcdir)/api/src\ + -DDATADIR=\"$(localstatedir)\" \ + -DCONFDIR=\"$(sysconfdir)/glusterfs\" \ + -DGSYNCD_PREFIX=\"$(libexecdir)/glusterfs\"\ + -DSYNCDAEMON_COMPILE=$(SYNCDAEMON_COMPILE) -DSBIN_DIR=\"$(sbindir)\"\ + $(XML_CPPFLAGS) + +AM_CFLAGS = -Wall $(GF_GLUSTERFS_CFLAGS) + +CLEANFILES = + +$(top_builddir)/libglusterfs/src/libglusterfs.la: + $(MAKE) -C $(top_builddir)/libglusterfs/src/ all diff --git a/heal/src/glfs-heal.c b/heal/src/glfs-heal.c new file mode 100644 index 00000000000..9e2be12ab4c --- /dev/null +++ b/heal/src/glfs-heal.c @@ -0,0 +1,488 @@ +/* + Copyright (c) 2014 Red Hat, Inc. <http://www.redhat.com> + This file is part of GlusterFS. + + This file is licensed to you under your choice of the GNU Lesser + General Public License, version 3 or any later version (LGPLv3 or + later), or the GNU General Public License, version 2 (GPLv2), in all + cases as published by the Free Software Foundation. +*/ + +#include <stdio.h> +#include <stdlib.h> +#include <errno.h> +#include "glfs.h" +#include "glfs-handles.h" +#include "glfs-internal.h" +#include "syncop.h" +#include <string.h> +#include <time.h> + +int +glfsh_link_inode_update_loc (loc_t *loc, struct iatt *iattr) +{ + inode_t *link_inode = NULL; + int ret = -1; + + link_inode = inode_link (loc->inode, NULL, NULL, iattr); + if (link_inode == NULL) + goto out; + + inode_unref (loc->inode); + loc->inode = link_inode; + ret = 0; +out: + return ret; +} + +int +glfsh_get_index_dir_loc (loc_t *rootloc, xlator_t *xl, loc_t *dirloc, + int32_t *op_errno) +{ + void *index_gfid = NULL; + int ret = 0; + dict_t *xattr = NULL; + struct iatt iattr = {0}; + struct iatt parent = {0}; + + ret = syncop_getxattr (xl, rootloc, &xattr, GF_XATTROP_INDEX_GFID); + if (ret < 0) { + *op_errno = errno; + goto out; + } + + ret = dict_get_ptr (xattr, GF_XATTROP_INDEX_GFID, &index_gfid); + if (ret < 0) { + *op_errno = EINVAL; + goto out; + } + + uuid_copy (dirloc->gfid, index_gfid); + dirloc->path = ""; + dirloc->inode = inode_new (rootloc->inode->table); + ret = syncop_lookup (xl, dirloc, NULL, + &iattr, NULL, &parent); + dirloc->path = NULL; + if (ret < 0) { + *op_errno = errno; + goto out; + } + ret = glfsh_link_inode_update_loc (dirloc, &iattr); + if (ret) + goto out; + glfs_loc_touchup (dirloc); + + return 0; +out: + if (xattr) + dict_unref (xattr); + return ret; +} + +static int +_set_self_heal_vxattrs (dict_t *xattr_req) +{ + int ret = 0; + + ret = dict_set_int32 (xattr_req, "dry-run-self-heal", 1); + if (ret < 0) + goto out; + + ret = dict_set_int32 (xattr_req, "attempt-self-heal", 1); + if (ret < 0) + goto out; + + ret = dict_set_int32 (xattr_req, "foreground-self-heal", 1); + if (ret < 0) + goto out; +out: + return ret; +} + +static gf_boolean_t +_is_self_heal_pending (dict_t *xattr_rsp) +{ + int ret = 0; + int sh_pending = 0; + + ret = dict_get_int32 (xattr_rsp, "metadata-self-heal-pending", + &sh_pending); + if ((ret == 0) && sh_pending) { + return _gf_true; + } + + ret = dict_get_int32 (xattr_rsp, "entry-self-heal-pending", + &sh_pending); + if ((ret == 0) && sh_pending) { + return _gf_true; + } + + ret = dict_get_int32 (xattr_rsp, "data-self-heal-pending", + &sh_pending); + if ((ret == 0) && sh_pending) { + return _gf_true; + } + return _gf_false; +} + +#define RESET_ENTRIES(loc, shf, ope, rsp, grsp) \ + do { \ + loc_wipe (&loc); \ + shf = 0; \ + ope = 0; \ + if (rsp) { \ + dict_unref (rsp); \ + rsp = NULL; \ + } \ + if (grsp) { \ + dict_unref (grsp); \ + grsp = NULL; \ + } \ + } while (0) + +static int +glfsh_build_index_loc (loc_t *loc, char *name, loc_t *parent) +{ + int ret = 0; + + uuid_copy (loc->pargfid, parent->inode->gfid); + loc->path = ""; + loc->name = name; + loc->parent = inode_ref (parent->inode); + if (!loc->parent) { + loc->path = NULL; + loc_wipe (loc); + ret = -1; + } + return ret; +} + +void +glfsh_remove_stale_index (xlator_t *xl, loc_t *parent, char *fname) +{ + int ret = 0; + loc_t index_loc = {0}; + + ret = glfsh_build_index_loc (&index_loc, fname, parent); + if (ret) + goto out; + ret = syncop_unlink (xl, &index_loc); + index_loc.path = NULL; + loc_wipe (&index_loc); +out: + return; +} +static int +glfsh_process_entries (xlator_t *xl, loc_t *parentloc, gf_dirent_t *entries, + off_t *offset, uint64_t *num_entries) +{ + gf_dirent_t *entry = NULL; + gf_dirent_t *tmp = NULL; + int ret = 0; + loc_t entry_loc = {0}; + struct iatt iattr = {0}; + struct iatt parent = {0}; + dict_t *xattr_req = NULL; + dict_t *xattr_rsp = NULL; + dict_t *getxattr_rsp = NULL; + int32_t sh_failed = 0; + int32_t sh_pending = 0; + char *path = NULL; + int32_t op_errno = 0; + + xattr_req = dict_new (); + if (!xattr_req) + goto out; + + list_for_each_entry_safe (entry, tmp, &entries->list, list) { + *offset = entry->d_off; + if ((strcmp (entry->d_name, ".") == 0) || + (strcmp (entry->d_name, "..") == 0)) + continue; + + RESET_ENTRIES (entry_loc, sh_failed, op_errno, xattr_rsp, + getxattr_rsp); + + ret = _set_self_heal_vxattrs (xattr_req); + if (ret < 0) + goto out; + + entry_loc.inode = inode_new (parentloc->inode->table); + if (!entry_loc.inode) + goto out; + + uuid_parse (entry->d_name, entry_loc.gfid); + //TODO: put gfid-path + glfs_loc_touchup (&entry_loc); + ret = syncop_lookup (xl->parents->xlator, &entry_loc, xattr_req, + &iattr, &xattr_rsp, &parent); + if (ret < 0) + op_errno = errno; + + if (op_errno == ENOENT || op_errno == ESTALE) { + glfsh_remove_stale_index (xl, parentloc, entry->d_name); + continue; + } + + ret = dict_get_int32 (xattr_rsp, "sh-failed", &sh_failed); + + sh_pending = _is_self_heal_pending (xattr_rsp); + //File/dir is undergoing I/O + if (!op_errno && !sh_failed && !sh_pending) + continue; + + ret = syncop_getxattr (xl, &entry_loc, &getxattr_rsp, + GFID_TO_PATH_KEY); + + ret = dict_get_str (getxattr_rsp, GFID_TO_PATH_KEY, &path); + + + (*num_entries)++; + printf ("%s\n", path ? path : uuid_utoa (entry_loc.gfid)); + } + ret = 0; +out: + RESET_ENTRIES (entry_loc, sh_failed, op_errno, xattr_rsp, getxattr_rsp); + return ret; +} + +static int +glfsh_crawl_directory (xlator_t *readdir_xl, fd_t *fd, loc_t *loc) +{ + off_t offset = 0; + gf_dirent_t entries; + int ret = 0; + gf_boolean_t free_entries = _gf_false; + uint64_t num_entries = 0; + + INIT_LIST_HEAD (&entries.list); + + while (1) { + ret = syncop_readdir (readdir_xl, fd, 131072, offset, &entries); + if (ret <= 0) + break; + ret = 0; + free_entries = _gf_true; + + if (list_empty (&entries.list)) + goto out; + + ret = glfsh_process_entries (readdir_xl, loc, &entries, &offset, + &num_entries); + if (ret < 0) + goto out; + + gf_dirent_free (&entries); + free_entries = _gf_false; + } + ret = 0; +out: + if (free_entries) + gf_dirent_free (&entries); + if (ret < 0) { + printf ("Failed to complete gathering info. " + "Number of entries so far: %"PRIu64"\n", num_entries); + } else { + printf ("Number of entries: %"PRIu64"\n", num_entries); + } + return ret; +} + +static int +glfsh_print_brick (xlator_t *xl, loc_t *rootloc) +{ + int ret = 0; + dict_t *xattr = NULL; + char *pathinfo = NULL; + char *brick_start = NULL; + char *brick_end = NULL; + + ret = syncop_getxattr (xl, rootloc, &xattr, GF_XATTR_PATHINFO_KEY); + if (ret < 0) + goto out; + + ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &pathinfo); + if (ret < 0) + goto out; + + brick_start = strchr (pathinfo, ':') + 1; + brick_end = pathinfo + strlen (pathinfo) - 1; + *brick_end = 0; + printf ("Brick %s\n", brick_start); + +out: + if (xattr) + dict_unref (xattr); + return ret; +} + +void +glfsh_print_brick_from_xl (xlator_t *xl) +{ + char *remote_host = NULL; + char *remote_subvol = NULL; + int ret = 0; + + ret = dict_get_str (xl->options, "remote-host", &remote_host); + if (ret < 0) + goto out; + + ret = dict_get_str (xl->options, "remote-subvolume", &remote_subvol); + if (ret < 0) + goto out; +out: + if (ret < 0) { + printf ("Brick - Not able to get brick information\n"); + } else { + printf ("Brick %s:%s\n", remote_host, remote_subvol); + } +} + +void +glfsh_print_pending_heals (xlator_t *xl, loc_t *rootloc) +{ + int ret = 0; + loc_t dirloc = {0}; + fd_t *fd = NULL; + int32_t op_errno = 0; + + ret = glfsh_print_brick (xl, rootloc); + if (ret < 0) { + glfsh_print_brick_from_xl (xl); + printf ("Status: %s\n", strerror (errno)); + goto out; + } + + ret = glfsh_get_index_dir_loc (rootloc, xl, &dirloc, &op_errno); + if (ret < 0) { + if (op_errno == ESTALE || op_errno == ENOENT) + printf ("Number of entries: 0\n"); + goto out; + } + + fd = fd_anonymous (dirloc.inode); + ret = glfsh_crawl_directory (xl, fd, &dirloc); + if (fd) + fd_unref (fd); + if (ret < 0) + printf ("Failed to find entries with pending self-heal\n"); +out: + loc_wipe (&dirloc); + return; +} + +static int +glfsh_validate_replicate_volume (xlator_t *xl) +{ + gf_boolean_t valid_replicate = _gf_false; + gf_boolean_t has_client = _gf_false; + while (xl->next) + xl = xl->next; + + while (xl) { + if (strcmp (xl->type, "protocol/client") == 0) { + has_client = _gf_true; + if (strcmp (xl->parents->xlator->type, + "cluster/replicate") != 0) { + valid_replicate = _gf_false; + goto out; + } + } + + xl = xl->prev; + } + valid_replicate = _gf_true; +out: + if (has_client && valid_replicate) + return 0; + return -1; +} + +int +main (int argc, char **argv) +{ + glfs_t *fs = NULL; + int ret = 0; + char *volname = NULL; + xlator_t *top_subvol = NULL; + xlator_t *xl = NULL; + loc_t rootloc = {0}; + + if (argc != 2) { + printf ("Usage: %s <volname>\n", argv[0]); + ret = -1; + goto out; + } + volname = argv[1]; + + fs = glfs_new (volname); + if (!fs) { + ret = -1; + printf ("Not able to initialize volume '%s'\n", volname); + goto out; + } + + ret = glfs_set_volfile_server (fs, "tcp", "localhost", 24007); + + ret = glfs_init (fs); + if (ret < 0) { + ret = -1; + if (errno == ENOENT) { + printf ("Volume %s does not exist\n", volname); + } else { + printf ("%s: Not able to fetch volfile from " + "glusterd\n", volname); + } + goto out; + } + + sleep (2); + __glfs_entry_fs (fs); + top_subvol = glfs_active_subvol (fs); + if (!top_subvol) { + ret = -1; + if (errno == ENOTCONN) { + printf ("Volume %s is not started (Or) All the bricks " + "are not running.\n", volname); + } else { + printf ("%s: Not able to mount the volume, %s\n", + volname, strerror (errno)); + } + goto out; + } + + ret = glfsh_validate_replicate_volume (top_subvol); + if (ret < 0) { + printf ("Volume %s is not of type replicate\n", volname); + goto out; + } + rootloc.inode = inode_ref (top_subvol->itable->root); + glfs_loc_touchup (&rootloc); + + xl = top_subvol; + while (xl->next) + xl = xl->next; + + while (xl) { + if (strcmp (xl->type, "protocol/client") == 0) { + glfsh_print_pending_heals (xl, &rootloc); + printf("\n"); + } + + xl = xl->prev; + } + + loc_wipe (&rootloc); + glfs_subvol_done (fs, top_subvol); + glfs_fini (fs); + + return 0; +out: + if (fs && top_subvol) + glfs_subvol_done (fs, top_subvol); + loc_wipe (&rootloc); + if (fs) + glfs_fini (fs); + + return ret; +} |