sanlock-devel December 2023

sanlock-devel@lists.fedorahosted.org

1 participants
7 discussions

[sanlock] branch master updated: release 3.9.0
by pagure＠pagure.io 13 Dec '23

13 Dec '23

This is an automated email from the git hooks/post-receive script. teigland pushed a commit to branch master in repository sanlock. The following commit(s) were added to refs/heads/master by this push: new 23766f9 release 3.9.0 23766f9 is described below commit 23766f9119b7c06b463110aeb9b969a742f0fe23 Author: David Teigland <teigland(a)redhat.com> AuthorDate: Wed Dec 13 16:47:19 2023 -0600 release 3.9.0 --- VERSION | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/VERSION b/VERSION index 0cbfaed..a5c4c76 100644 --- a/VERSION +++ b/VERSION @@ -1 +1 @@ -3.8.5 +3.9.0 -- To stop receiving notification emails like this one, please contact the administrator of this repository.

1 0

[sanlock] 06/06: sanlock: replace strncat for creating debug output
by pagure＠pagure.io 13 Dec '23

13 Dec '23

This is an automated email from the git hooks/post-receive script. teigland pushed a commit to branch master in repository sanlock. commit 19575183392dab154ca1b9962189cd666c4576b6 Author: David Teigland <teigland(a)redhat.com> AuthorDate: Tue Dec 12 17:50:58 2023 -0600 sanlock: replace strncat for creating debug output --- src/paxos_lease.c | 35 ++++++++++++++++++++++++++++++++--- 1 file changed, 32 insertions(+), 3 deletions(-) diff --git a/src/paxos_lease.c b/src/paxos_lease.c index 179d8e7..5643739 100644 --- a/src/paxos_lease.c +++ b/src/paxos_lease.c @@ -51,6 +51,35 @@ int get_rand(int a, int b); #define BK_DEBUG_COUNT 4 #define BK_STR_SIZE 80 +static void bk_debug_append(char *bk_debug, char *bk_str) +{ + int off = 0; + int i, j; + + /* bk_debug buffer is BK_DEBUG_SIZE, bk_str is BK_STR_SIZE */ + /* include as much debugging text in bk_debug as will fit */ + + for (i = 0; i < BK_DEBUG_SIZE - 1; i++) { + if (!bk_debug[i]) { + off = i; + break; + } + } + + if (off >= (BK_DEBUG_SIZE-1)) { + bk_debug[BK_DEBUG_SIZE-1] = '\0'; + return; + } + + for (j = 0; j < strlen(bk_str); j++) { + bk_debug[off++] = bk_str[j]; + if (off >= BK_DEBUG_SIZE) { + bk_debug[BK_DEBUG_SIZE-1] = '\0'; + return; + } + } +} + static uint32_t roundup_power_of_two(uint32_t val) { val--; @@ -577,7 +606,7 @@ static int run_ballot(struct task *task, struct token *token, uint32_t flags, (unsigned long long)bk_in.lver, bk_in.flags); bk_str[BK_STR_SIZE-1] = '\0'; - strncat(bk_debug, bk_str, BK_STR_SIZE-1); + bk_debug_append(bk_debug, bk_str); bk_debug_count++; } @@ -790,7 +819,7 @@ static int run_ballot(struct task *task, struct token *token, uint32_t flags, (unsigned long long)bk->lver, bk->flags); bk_str[BK_STR_SIZE-1] = '\0'; - strncat(bk_debug, bk_str, BK_STR_SIZE-1); + bk_debug_append(bk_debug, bk_str); bk_debug_count++; } @@ -1407,7 +1436,7 @@ static int _lease_read_one(struct task *task, (unsigned long long)bk.lver, bk.flags); bk_str[BK_STR_SIZE-1] = '\0'; - strncat(bk_debug, bk_str, BK_STR_SIZE-1); + bk_debug_append(bk_debug, bk_str); bk_debug_count++; } -- To stop receiving notification emails like this one, please contact the administrator of this repository.

1 0

[sanlock] 05/06: sanlock: reduce warnings
by pagure＠pagure.io 13 Dec '23

13 Dec '23

This is an automated email from the git hooks/post-receive script. teigland pushed a commit to branch master in repository sanlock. commit 666d4da3769b91b40c7030190360e33e94f921ad Author: David Teigland <teigland(a)redhat.com> AuthorDate: Tue Dec 12 16:14:32 2023 -0600 sanlock: reduce warnings Stop treating the non-null-terminated buffers as strings to avoid warnings. --- src/cmd.c | 12 ++++++------ src/delta_lease.c | 2 +- src/direct.c | 12 ++++++------ src/leader.h | 4 ++-- src/lockspace.c | 9 +++++---- src/main.c | 11 +++++++---- src/paxos_lease.c | 4 ++-- src/resource.c | 4 ++-- src/rindex.c | 2 +- src/sanlock.h | 12 ++++++------ src/sanlock_internal.h | 4 ++-- 11 files changed, 40 insertions(+), 36 deletions(-) diff --git a/src/cmd.c b/src/cmd.c index ae57bf3..64ee409 100644 --- a/src/cmd.c +++ b/src/cmd.c @@ -358,7 +358,7 @@ static void cmd_acquire(struct task *task, struct cmd_args *ca, uint32_t cmd) goto done; } - strncpy(cl->owner_name, opt.owner_name, SANLK_NAME_LEN); + memcpy(cl->owner_name, opt.owner_name, SANLK_NAME_LEN); /* data after opt is not used */ /* @@ -2511,7 +2511,7 @@ static void send_state_daemon(int fd) int str_len; memset(&st, 0, sizeof(st)); - strncpy(st.name, our_host_name_global, NAME_ID_SIZE); + memcpy(st.name, our_host_name_global, NAME_ID_SIZE); st.type = SANLK_STATE_DAEMON; @@ -2534,7 +2534,7 @@ static void send_state_client(int fd, struct client *cl, int ci) st.type = SANLK_STATE_CLIENT; st.data32 = cl->pid; - strncpy(st.name, cl->owner_name, NAME_ID_SIZE); + memcpy(st.name, cl->owner_name, NAME_ID_SIZE); str_len = print_state_client(cl, ci, str); @@ -2556,7 +2556,7 @@ static void send_state_lockspace(int fd, struct space *sp, const char *list_name st.type = SANLK_STATE_LOCKSPACE; st.data64 = sp->host_id; - strncpy(st.name, sp->space_name, NAME_ID_SIZE); + memcpy(st.name, sp->space_name, NAME_ID_SIZE); str_len = print_state_lockspace(sp, str, list_name); @@ -2567,7 +2567,7 @@ static void send_state_lockspace(int fd, struct space *sp, const char *list_name send_all(fd, str, str_len, MSG_NOSIGNAL); memset(&lockspace, 0, sizeof(struct sanlk_lockspace)); - strncpy(lockspace.name, sp->space_name, NAME_ID_SIZE); + memcpy(lockspace.name, sp->space_name, NAME_ID_SIZE); lockspace.host_id = sp->host_id; memcpy(&lockspace.host_id_disk, &sp->host_id_disk, sizeof(struct sanlk_disk)); @@ -2590,7 +2590,7 @@ void send_state_resource(int fd, struct resource *r, const char *list_name, st.type = SANLK_STATE_RESOURCE; st.data32 = pid; st.data64 = r->leader.lver; - strncpy(st.name, r->r.name, NAME_ID_SIZE); + memcpy(st.name, r->r.name, NAME_ID_SIZE); str_len = print_state_resource(r, str, list_name, token_id); diff --git a/src/delta_lease.c b/src/delta_lease.c index d36d4e0..4ac63ef 100644 --- a/src/delta_lease.c +++ b/src/delta_lease.c @@ -900,7 +900,7 @@ int delta_lease_init(struct task *task, leader.max_hosts = 1; leader.timestamp = LEASE_FREE; leader.io_timeout = io_timeout; - strncpy(leader.space_name, ls->name, NAME_ID_SIZE); + memcpy(leader.space_name, ls->name, NAME_ID_SIZE); leader.checksum = 0; /* set below */ /* make the first record invalid so we can do a single atomic diff --git a/src/direct.c b/src/direct.c index f3ea7d1..73861a2 100644 --- a/src/direct.c +++ b/src/direct.c @@ -728,8 +728,8 @@ int direct_dump(struct task *task, char *dump_path, int force_mode) if (!lr->owner_id && !lr->owner_generation) continue; - strncpy(sname, lr->space_name, NAME_ID_SIZE); - strncpy(rname, lr->resource_name, NAME_ID_SIZE); + strcpy(sname, lr->space_name); + strcpy(rname, lr->resource_name); printf("%08llu %36s %48s %010llu %04llu %04llu", (unsigned long long)(start_offset + ((sector_nr + i) * sector_size)), @@ -752,8 +752,8 @@ int direct_dump(struct task *task, char *dump_path, int force_mode) leader_record_in(lr_end, &lr_in); lr = &lr_in; - strncpy(sname, lr->space_name, NAME_ID_SIZE); - strncpy(rname, lr->resource_name, NAME_ID_SIZE); + strcpy(sname, lr->space_name); + strcpy(rname, lr->resource_name); printf("%08llu %36s %48s %010llu %04llu %04llu %llu", (unsigned long long)(start_offset + (sector_nr * sector_size)), @@ -804,7 +804,7 @@ int direct_dump(struct task *task, char *dump_path, int force_mode) rindex_header_in(rh_end, &rh_in); rh = &rh_in; - strncpy(sname, rh->lockspace_name, NAME_ID_SIZE); + strcpy(sname, rh->lockspace_name); printf("%08llu %36s rindex_header 0x%x %d %u %llu\n", (unsigned long long)(start_offset + (sector_nr * sector_size)), @@ -872,7 +872,7 @@ int direct_next_free(struct task *task, char *path) sd.offset = atoll(off_str); } - strncpy(sd.path, path, SANLK_PATH_LEN); + strncpy(sd.path, path, SANLK_PATH_LEN-1); sd.fd = -1; rv = open_disk(&sd); diff --git a/src/leader.h b/src/leader.h index 7656299..649bcf8 100644 --- a/src/leader.h +++ b/src/leader.h @@ -69,8 +69,8 @@ struct leader_record { uint64_t owner_id; /* host_id of owner */ uint64_t owner_generation; uint64_t lver; - char space_name[NAME_ID_SIZE]; /* lockspace for resource */ - char resource_name[NAME_ID_SIZE]; /* resource being locked */ + char space_name[NAME_ID_SIZE] __attribute__ ((nonstring)); /* lockspace for resource */ + char resource_name[NAME_ID_SIZE] __attribute__ ((nonstring)); /* resource being locked */ uint64_t timestamp; uint64_t unused1; uint32_t checksum; diff --git a/src/lockspace.c b/src/lockspace.c index a874148..da51af0 100644 --- a/src/lockspace.c +++ b/src/lockspace.c @@ -449,7 +449,7 @@ void check_other_leases(struct space *sp, char *buf) (unsigned long long)leader->owner_generation, (unsigned long long)leader->timestamp, leader->resource_name); - strncpy(hs->owner_name, leader->resource_name, NAME_ID_SIZE); + memcpy(hs->owner_name, leader->resource_name, NAME_ID_SIZE); } if (hs->owner_id == leader->owner_id && @@ -469,7 +469,7 @@ void check_other_leases(struct space *sp, char *buf) if (!hs->lease_bad) { hs->owner_id = leader->owner_id; hs->owner_generation = leader->owner_generation; - strncpy(hs->owner_name, leader->resource_name, NAME_ID_SIZE); + memcpy(hs->owner_name, leader->resource_name, NAME_ID_SIZE); hs->io_timeout = leader->io_timeout; } @@ -1057,9 +1057,10 @@ int add_lockspace_start(struct sanlk_lockspace *ls, uint32_t io_timeout, struct int i; if (!ls->name[0] || !ls->host_id || !ls->host_id_disk.path[0]) { - log_error("add_lockspace bad args id %llu name %zu path %zu", + log_error("add_lockspace bad args host_id %llu name %s path %s", (unsigned long long)ls->host_id, - strlen(ls->name), strlen(ls->host_id_disk.path)); + ls->name[0] ? "set" : "empty", + ls->host_id_disk.path[0] ? "set" : "empty"); return -EINVAL; } diff --git a/src/main.c b/src/main.c index 4e7f3c0..96ad50a 100644 --- a/src/main.c +++ b/src/main.c @@ -1904,8 +1904,11 @@ static int parse_arg_rentry(char *str) com.rentry.offset = offnum; } - if (name) - strncpy(com.rentry.name, name, SANLK_NAME_LEN); + if (name) { + char tmps[SANLK_NAME_LEN+1] = { 0 }; + strncpy(tmps, name, SANLK_NAME_LEN); + memcpy(com.rentry.name, tmps, SANLK_NAME_LEN); + } return 0; } @@ -2886,7 +2889,7 @@ static void read_config_file(void) } else if (!strcmp(str, "our_host_name")) { memset(str, 0, sizeof(str)); get_val_str(line, str); - strncpy(com.our_host_name, str, NAME_ID_SIZE); + memcpy(com.our_host_name, str, NAME_ID_SIZE); } else if (!strcmp(str, "renewal_read_extend_sec")) { /* zero is a valid setting so we need the _set field to say it's set */ @@ -3319,7 +3322,7 @@ static int do_client(void) if (!res_ls) break; memset(res_ls, 0, sizeof(struct sanlk_resource)); - strcpy(res_ls->lockspace_name, com.lockspace.name); + memcpy(res_ls->lockspace_name, com.lockspace.name, SANLK_NAME_LEN); com.res_args[0] = res_ls; com.res_count = 1; } diff --git a/src/paxos_lease.c b/src/paxos_lease.c index ce265e7..179d8e7 100644 --- a/src/paxos_lease.c +++ b/src/paxos_lease.c @@ -2467,8 +2467,8 @@ int paxos_lease_init(struct task *task, leader.sector_size = sector_size; leader.num_hosts = num_hosts; leader.max_hosts = max_hosts; - strncpy(leader.space_name, token->r.lockspace_name, NAME_ID_SIZE); - strncpy(leader.resource_name, token->r.name, NAME_ID_SIZE); + memcpy(leader.space_name, token->r.lockspace_name, NAME_ID_SIZE); + memcpy(leader.resource_name, token->r.name, NAME_ID_SIZE); leader.checksum = 0; /* set after leader_record_out */ memset(&rr, 0, sizeof(rr)); diff --git a/src/resource.c b/src/resource.c index 0235b53..c1c49f9 100644 --- a/src/resource.c +++ b/src/resource.c @@ -107,9 +107,9 @@ static struct resource *get_free_resource(struct token *token, int *token_matche /* find a previous r that matches token */ list_for_each_entry(r, &resources_free, list) { - if (strcmp(r->r.lockspace_name, token->r.lockspace_name)) + if (memcmp(r->r.lockspace_name, token->r.lockspace_name, SANLK_NAME_LEN)) continue; - if (strcmp(r->r.name, token->r.name)) + if (memcmp(r->r.name, token->r.name, SANLK_NAME_LEN)) continue; if (r->r.num_disks != token->r.num_disks) continue; diff --git a/src/rindex.c b/src/rindex.c index 9ef02da..0ac503e 100644 --- a/src/rindex.c +++ b/src/rindex.c @@ -437,7 +437,7 @@ int rindex_format(struct task *task, struct sanlk_rindex *ri) rh.sector_size = sector_size; rh.max_resources = max_resources; rh.rx_offset = rx.disk->offset; - strncpy(rh.lockspace_name, rx.ri->lockspace_name, NAME_ID_SIZE); + memcpy(rh.lockspace_name, rx.ri->lockspace_name, NAME_ID_SIZE); memset(&rh_end, 0, sizeof(struct rindex_header)); rindex_header_out(&rh, &rh_end); diff --git a/src/sanlock.h b/src/sanlock.h index 5999630..978aac7 100644 --- a/src/sanlock.h +++ b/src/sanlock.h @@ -96,8 +96,8 @@ struct sanlk_disk { #define SANLK_RES_SECTOR4K 0x00000200 struct sanlk_resource { - char lockspace_name[SANLK_NAME_LEN]; /* terminating \0 not required */ - char name[SANLK_NAME_LEN]; /* terminating \0 not required */ + char lockspace_name[SANLK_NAME_LEN] __attribute__ ((nonstring)); /* terminating \0 not required */ + char name[SANLK_NAME_LEN] __attribute__ ((nonstring)); /* terminating \0 not required */ uint64_t lver; /* use with SANLK_RES_LVER */ uint64_t data64; /* per-resource command-specific data */ uint32_t data32; /* per-resource command-specific data */ @@ -120,12 +120,12 @@ struct sanlk_rindex { uint32_t flags; /* SANLK_RIF_ */ uint32_t max_resources; /* the max res structs that will follow rindex */ uint64_t unused; - char lockspace_name[SANLK_NAME_LEN]; /* terminating \0 not required */ + char lockspace_name[SANLK_NAME_LEN] __attribute__ ((nonstring)); /* terminating \0 not required */ struct sanlk_disk disk; /* location of rindex */ }; struct sanlk_rentry { - char name[SANLK_NAME_LEN]; /* terminating \0 not required */ + char name[SANLK_NAME_LEN] __attribute__ ((nonstring)); /* terminating \0 not required */ uint64_t offset; uint32_t flags; uint32_t unused; @@ -135,7 +135,7 @@ struct sanlk_rentry { that requires the extra work of segmenting it by resource name) */ struct sanlk_options { - char owner_name[SANLK_NAME_LEN]; /* optional user friendly name */ + char owner_name[SANLK_NAME_LEN] __attribute__ ((nonstring)); /* optional user friendly name */ uint32_t flags; uint32_t len; /* unused and ignored */ /* followed by len bytes (unused and ignored) */ @@ -154,7 +154,7 @@ struct sanlk_options { #define SANLK_LSF_SECTOR4K 0x00000200 struct sanlk_lockspace { - char name[SANLK_NAME_LEN]; + char name[SANLK_NAME_LEN] __attribute__ ((nonstring)); uint64_t host_id; uint32_t flags; /* SANLK_LSF_ */ struct sanlk_disk host_id_disk; diff --git a/src/sanlock_internal.h b/src/sanlock_internal.h index d8c027f..3e06e4f 100644 --- a/src/sanlock_internal.h +++ b/src/sanlock_internal.h @@ -179,7 +179,7 @@ struct host_status { uint64_t set_bit_time; uint16_t io_timeout; uint16_t lease_bad; - char owner_name[NAME_ID_SIZE]; + char owner_name[NAME_ID_SIZE] __attribute__ ((nonstring)); }; struct renewal_history { @@ -198,7 +198,7 @@ struct renewal_history { struct space { struct list_head list; - char space_name[NAME_ID_SIZE]; + char space_name[NAME_ID_SIZE] __attribute__ ((nonstring)); uint32_t space_id; /* used to refer to this space instance in log messages */ uint64_t host_id; uint64_t host_generation; -- To stop receiving notification emails like this one, please contact the administrator of this repository.

1 0

[sanlock] 04/06: sanlock: fix direct_next_free missing init
by pagure＠pagure.io 13 Dec '23

13 Dec '23

This is an automated email from the git hooks/post-receive script. teigland pushed a commit to branch master in repository sanlock. commit 4a9f303429a01a0dda441a03c1c08fb9ac5dd5da Author: David Teigland <teigland(a)redhat.com> AuthorDate: Tue Dec 12 13:52:51 2023 -0600 sanlock: fix direct_next_free missing init --- src/direct.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/direct.c b/src/direct.c index 7e00c52..f3ea7d1 100644 --- a/src/direct.c +++ b/src/direct.c @@ -860,7 +860,7 @@ int direct_next_free(struct task *task, char *path) struct leader_record lr; struct sync_disk sd; uint64_t sector_nr; - int sector_size, sector_count, datalen, align_size; + int sector_size = 0, sector_count, datalen, align_size = 0; int rv; memset(&sd, 0, sizeof(struct sync_disk)); -- To stop receiving notification emails like this one, please contact the administrator of this repository.

1 0

[sanlock] 03/06: sanlock: use product_uuid for host name
by pagure＠pagure.io 13 Dec '23

13 Dec '23

This is an automated email from the git hooks/post-receive script. teigland pushed a commit to branch master in repository sanlock. commit d98b217dee02f436c1c3fe99e2c87b92a1019d9d Author: David Teigland <teigland(a)redhat.com> AuthorDate: Mon Dec 11 17:16:32 2023 -0600 sanlock: use product_uuid for host name When there is no host name specified on the command line or in the config file, use the value from /sys/devices/virtual/dmi/id/product_uuid instead of generating a random UUID each time the daemon is started. If no product_uuid exists, the random UUID will be generated for the host name. --- src/main.c | 51 +++++++++++++++++++++++++++++++++++++++------------ src/sanlock.8 | 6 +++--- 2 files changed, 42 insertions(+), 15 deletions(-) diff --git a/src/main.c b/src/main.c index a9ceee5..4e7f3c0 100644 --- a/src/main.c +++ b/src/main.c @@ -1473,37 +1473,64 @@ int get_rand(int a, int b) return a + (int) (((float)(b - a + 1)) * rv / (RAND_MAX+1.0)); } +static void read_product_uuid(char *buf, size_t buf_size) +{ + FILE *fp; + size_t len; + + if (!(fp = fopen("/sys/devices/virtual/dmi/id/product_uuid", "r"))) + return; + + if (!fgets(buf, buf_size, fp)) + return; + + if ((len = strlen(buf)) && buf[len - 1] == '\n') + buf[--len] = '\0'; + + fclose(fp); +} + static void setup_host_name(void) { - struct utsname name; + char product_uuid[SANLK_NAME_LEN+1] = { 0 }; char uuid[37]; + struct utsname name; uuid_t uu; int ret; - memset(rand_state, 0, sizeof(rand_state)); - initstate(time(NULL), rand_state, sizeof(rand_state)); - - /* use host name from command line */ + /* + * Get host name value from: + * 1. command line + * 2. config file + * 3. /sys/devices/virtual/dmi/id/product_uuid + * 4. generate random uuid that won't collide with another host + */ if (com.our_host_name[0]) { memcpy(our_host_name_global, com.our_host_name, SANLK_NAME_LEN); return; } - /* make up something that's likely to be different among hosts */ - memset(&our_host_name_global, 0, sizeof(our_host_name_global)); + memset(product_uuid, 0, sizeof(product_uuid)); memset(&name, 0, sizeof(name)); memset(&uuid, 0, sizeof(uuid)); + read_product_uuid(product_uuid, sizeof(product_uuid)); uname(&name); - uuid_generate(uu); - uuid_unparse_lower(uu, uuid); - ret = snprintf(our_host_name_global, NAME_ID_SIZE, "%s.", uuid); + if (product_uuid[0]) { + ret = snprintf(our_host_name_global, SANLK_NAME_LEN, "%s.", product_uuid); + } else { + memset(rand_state, 0, sizeof(rand_state)); + initstate(time(NULL), rand_state, sizeof(rand_state)); + uuid_generate(uu); + uuid_unparse_lower(uu, uuid); + ret = snprintf(our_host_name_global, SANLK_NAME_LEN, "%s.", uuid); + } - if (ret < NAME_ID_SIZE) - memcpy(our_host_name_global+ret, name.nodename, NAME_ID_SIZE-ret); + if (ret < SANLK_NAME_LEN) + memcpy(our_host_name_global+ret, name.nodename, SANLK_NAME_LEN-ret); } static void setup_limits(void) diff --git a/src/sanlock.8 b/src/sanlock.8 index 0776b8f..5295bf5 100644 --- a/src/sanlock.8 +++ b/src/sanlock.8 @@ -120,9 +120,9 @@ delta lease disk sector, reading it back after a delay, and verifying it is the same. .IP \[bu] -If a unique host name is not specified, sanlock generates a uuid to use -as the host's name. The delta lease algorithm depends on hosts using -unique names. +If a unique host name is not specified, sanlock uses the product_uuid +if one is available, otherwise generates a uuid to use as the host's +name. The delta lease algorithm depends on hosts using unique names. .IP \[bu] The application on each host should be configured with a unique host_id, -- To stop receiving notification emails like this one, please contact the administrator of this repository.

1 0

[sanlock] 02/06: sanlock: add comment with timeout example
by pagure＠pagure.io 13 Dec '23

13 Dec '23

This is an automated email from the git hooks/post-receive script. teigland pushed a commit to branch master in repository sanlock. commit d737809488521c15bc60fac9e34a7b872d2ac94c Author: David Teigland <teigland(a)redhat.com> AuthorDate: Mon Dec 11 16:36:27 2023 -0600 sanlock: add comment with timeout example --- src/timeouts.h | 77 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 77 insertions(+) diff --git a/src/timeouts.h b/src/timeouts.h index ac9842a..73b02f1 100644 --- a/src/timeouts.h +++ b/src/timeouts.h @@ -464,6 +464,83 @@ * watchdog really fires. */ +/* + * Example of how long it takes to acquire a lease that was held + * by a failed host. + * + * wdmd test interval = 5 + * watchdog_fire_timeout = 30 + * io_timeout_seconds = 10 + * id_renewal_seconds = 20 + * id_renewal_fail_seconds = 80 + * host_dead_seconds = 110 (id_renewal_fail_seconds + watchdog_fire_timeout) + * + * host that was killed + * -------------------- + * sanlock debug output: + * 13:31:44 7630 [2157]: s1 delta_renew begin read + * 13:31:44 7630 [2157]: s1 delta_renew begin write for new ts 7630 + * 13:31:44 7630 [2157]: s1 renewed 7630 delta_length 0 interval 20 + * <host killed> + * + * wdmd debug output: + * 1702323101 keepalive 0 + * 1702323101 test_interval 5 sleep_seconds 5 poll_timeout 5000 + * 1702323104 test_live ci 2 renewal 7630 expire 7710 + * 1702323104 test_interval 5 sleep_seconds 3 poll_timeout 3000 + * <host killed> + * + * host_dead_seconds (110) after last successful renewal (13:31:44) + * would be 13:33:34 as the earliest another host can take a lease. + * + * host that was trying to acquire lease held by dead host + * ------------------------------------------------------- + * running: while true; do date; lvchange -ay bb/lvol1; sleep 1; done + * Mon Dec 11 13:33:50 CST 2023 + * LV locked by other host: bb/lvol1 + * Failed to lock logical volume bb/lvol1. + * Mon Dec 11 13:33:51 CST 2023 + * <success> + * + * sanlock debug output: + * 13:31:40 7584 [2342]: s1 renewed 7584 delta_length 0 interval 20 + * ... + * 13:32:00 7605 [2342]: s1 renewed 7605 delta_length 0 interval 21 + * ... + * in that renewal at 7605, we see the last renewal from the other host, + * which is shown above from the other host's logs at "13:31:44 7630". + * we record our own current time 7605 as the last live time for the dead host. + * our initial attempts to get a lease held by the dead host will fail + * as follows until host_dead_seconds has elapased: + * paxos_acquire owner 2 2 7431 host_status 2 2 7630 wait_start 7605 + * paxos_acquire owner 2 2 7431 no wait + * acquire_disk rv -247 lver 0 at 0 + * acquire_token disk error -247 + * release_token r_flags 8 lver 0 + * release_token done r_flags 8 + * cmd_acquire 2,9,2124 acquire_token -247 lease owned by other host + * + * after retrying for host_dead_seconds, it succeeds: + * 13:33:51 7716 [2282]: s1:r5 paxos_acquire begin offset 71303168 0xa 512 1048576 + * 13:33:51 7716 [2282]: s1:r5 leader 2 owner 2 2 7431 dblocks 1:2002:2002:2:2:7431:2:0, + * 13:33:51 7716 [2282]: s1:r5 paxos_acquire leader 2 owner 2 2 7431 max mbal[1] 2002 our_dblock 0 0 0 0 0 0 + * 13:33:51 7716 [2282]: s1:r5 paxos_acquire owner 2 2 7431 host_status 2 2 7630 wait_start 7605 + * 13:33:51 7716 [2282]: s1:r5 paxos_acquire owner 2 2 7431 delta 2 2 7630 dead 7716-7605>110 + * 13:33:51 7716 [2282]: s1:r5 paxos_acquire leader 2 owner 2 2 7431 + * + * This shows that the difference between the current time (7716) and the + * time it last saw the dead host alive (7605) is >= host_dead_seconds (110), + * so it's allowed to get the lease. + * + * The host was killed at 13:31:44 and its lease was acquired at 13:33:51. + * This is 127 seconds, which includes the 110 host_dead_seconds, plus + * 17 extra seconds that elapsed between the time the host was killed + * and the time that the other host ran its own renewal and recorded the + * final renewal of the dead host. + * The host that was killed last renewed at 13:31:44, and the other host's + * next renewal after that was at 13:32:00 which is a 16 second difference. + */ + /* * killing pids -- To stop receiving notification emails like this one, please contact the administrator of this repository.

1 0

[sanlock] 01/06: sanlock: skip short delay in delta lease reacquire
by pagure＠pagure.io 13 Dec '23

13 Dec '23

This is an automated email from the git hooks/post-receive script. teigland pushed a commit to branch master in repository sanlock. commit a239c898c12a92cb5fb74f0eb74bc5919db45b94 Author: David Teigland <teigland(a)redhat.com> AuthorDate: Fri Dec 8 17:48:02 2023 -0600 sanlock: skip short delay in delta lease reacquire When a host with the same host name reacquires the same host_id, the "fast" delta lease reacquire was still applying the short delay (default 20s) which should not be necessary. Reduce this further to just 1 second when the lease is also free. --- src/delta_lease.c | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-) diff --git a/src/delta_lease.c b/src/delta_lease.c index 9d5aafd..d36d4e0 100644 --- a/src/delta_lease.c +++ b/src/delta_lease.c @@ -340,6 +340,7 @@ int delta_lease_acquire(struct task *task, uint32_t checksum; int other_io_timeout, other_host_dead_seconds, other_id_renewal_seconds; int i, error, rv, delay, delta_large_delay; + int fast_free_delay = 0; log_space(sp, "delta_acquire begin %.48s:%llu", sp->space_name, (unsigned long long)host_id); @@ -361,6 +362,19 @@ int delta_lease_acquire(struct task *task, leader.io_timeout, sp->io_timeout); } + /* + * This delay-free reacquire could probably be used whenever + * our_host_name matches the current leader.resource_name, + * regardless of wheter the lease is free (i.e. we cleanly + * released the lease last time we held it.) + */ + if (!strncmp(leader.resource_name, our_host_name, NAME_ID_SIZE) && + (leader.timestamp == LEASE_FREE)) { + log_space(sp, "delta_acquire free fast reacquire"); + fast_free_delay = 1; + goto write_new; + } + if (leader.timestamp == LEASE_FREE) goto write_new; @@ -475,7 +489,10 @@ int delta_lease_acquire(struct task *task, memcpy(&leader1, &leader, sizeof(struct leader_record)); - delay = 2 * other_io_timeout; + if (fast_free_delay) + delay = 1; + else + delay = 2 * other_io_timeout; log_space(sp, "delta_acquire delta_short_delay %d", delay); for (i = 0; i < delay; i++) { -- To stop receiving notification emails like this one, please contact the administrator of this repository.

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

sanlock-devel December 2023