cluster-commits February 2012

cluster-commits@lists.stg.fedorahosted.org

7 participants
60 discussions

cluster: RHEL56 - cman: Clean shutdown_con if the controlling process is killed.
by Fabio M. Di Nitto 27 Feb '12

27 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=16… Commit: 1671c39acedc665dc827165d0aa0a1c696c2da8f Parent: 36fd94d68003389d9585c87a649726742e06badd Author: Christine Caulfield <ccaulfie(a)redhat.com> AuthorDate: Fri Sep 19 13:02:40 2008 +0100 Committer: Fabio M. Di Nitto <fdinitto(a)redhat.com> CommitterDate: Mon Feb 27 15:54:57 2012 +0100 cman: Clean shutdown_con if the controlling process is killed. If a shutdown is initiated by a process that is then killed, the shutdown_con isn't cleared. So if another process replies to the shutdown request cman could segfault. Resolves: rhbz#795814 Signed-off-by: Christine Caulfield <ccaulfie(a)redhat.com> Signed-off-by: Fabio M. Di Nitto <fdintto(a)redhat.com> --- cman/daemon/commands.c | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-) diff --git a/cman/daemon/commands.c b/cman/daemon/commands.c index 6ffe9e5..781e564 100644 --- a/cman/daemon/commands.c +++ b/cman/daemon/commands.c @@ -1482,6 +1482,11 @@ void unbind_con(struct connection *con) check_shutdown_status(); } + + /* If the controlling shutdown process has quit, then cancel the + shutdown session */ + if (con == shutdown_con) + shutdown_con = NULL; } /* Post a PORT OPEN/CLOSE event to anyone listening on this end */

1 0

dlm: master - dlm_controld: fix stonith calls
by David Teigland 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=ee… Commit: eeb80cd73160856a16f73fc45d1ab48904aa55f0 Parent: 2ad89c869cbd6338361ccf25f7c84c719d88e5a9 Author: David Teigland <teigland(a)redhat.com> AuthorDate: Wed Feb 22 14:37:52 2012 -0600 Committer: David Teigland <teigland(a)redhat.com> CommitterDate: Wed Feb 22 14:38:56 2012 -0600 dlm_controld: fix stonith calls and handle errors Signed-off-by: David Teigland <teigland(a)redhat.com> --- dlm_controld/cpg.c | 21 +++++++++++++++------ dlm_controld/dlm_daemon.h | 2 +- dlm_controld/fence.c | 20 +++++++++----------- 3 files changed, 25 insertions(+), 18 deletions(-) diff --git a/dlm_controld/cpg.c b/dlm_controld/cpg.c index ced912f..1360493 100644 --- a/dlm_controld/cpg.c +++ b/dlm_controld/cpg.c @@ -618,8 +618,10 @@ static int check_fencing_done(struct lockspace *ls) fenced since node->start_time */ rv = fence_node_time(node->nodeid, &last_fenced_time); - if (rv < 0) - log_error("fenced_node_info error %d", rv); + if (rv < 0) { + log_error("fenced_node_time error %d", rv); + continue; + } /* fenced gives us real time */ @@ -653,8 +655,10 @@ static int check_fencing_done(struct lockspace *ls) } } - if (wait_count) + if (wait_count) { + log_group(ls, "check_fencing wait_count %d", wait_count); return 0; + } /* now check if there are any outstanding fencing ops (for nodes we may not have seen in any lockspace), and return 0 if there @@ -666,8 +670,10 @@ static int check_fencing_done(struct lockspace *ls) return 0; } - if (in_progress) + if (in_progress) { + log_group(ls, "check_fencing in progress %d", in_progress); return 0; + } log_group(ls, "check_fencing done"); return 1; @@ -694,12 +700,15 @@ static int need_fencing(struct lockspace *ls) static void request_fencing(struct lockspace *ls) { struct node *node; + int rv; list_for_each_entry(node, &ls->node_history, list) { if (!node->request_fencing) continue; - fence_request(node->nodeid); - node->request_fencing = 0; + log_group(ls, "fence_request %d", node->nodeid); + rv = fence_request(node->nodeid); + if (!rv) + node->request_fencing = 0; } } diff --git a/dlm_controld/dlm_daemon.h b/dlm_controld/dlm_daemon.h index 3899550..5fca041 100644 --- a/dlm_controld/dlm_daemon.h +++ b/dlm_controld/dlm_daemon.h @@ -348,7 +348,7 @@ void process_cluster_cfg(int ci); void kick_node_from_cluster(int nodeid); /* fence.c */ -void fence_request(int nodeid); +int fence_request(int nodeid); int fence_node_time(int nodeid, uint64_t *last_fenced_time); int fence_in_progress(int *count); diff --git a/dlm_controld/fence.c b/dlm_controld/fence.c index f91d9de..50b7d4d 100644 --- a/dlm_controld/fence.c +++ b/dlm_controld/fence.c @@ -7,30 +7,28 @@ */ #include "dlm_daemon.h" -#ifdef STONITH #include <pacemaker/crm/stonith-ng.h> -#endif -void fence_request(int nodeid) +int fence_request(int nodeid) { -#ifdef STONITH int rv; - rv = stonith_api_kick_cs_helper(nodeid, 300, 1); - if (rv) - log_error("stonith_api_kick_cs_helper %d error %d", nodeid, rv); -#endif + rv = stonith_api_kick_helper(nodeid, 300, 1); + if (rv) { + log_error("stonith_api_kick_helper %d error %d", nodeid, rv); + return rv; + } + return 0; } int fence_node_time(int nodeid, uint64_t *last_fenced_time) { -#ifdef STONITH - *last_fenced_time = stonith_api_time_cs_helper(nodeid, 0); -#endif + *last_fenced_time = stonith_api_time_helper(nodeid, 0); return 0; } int fence_in_progress(int *count) { + *count = 0; return 0; }

1 0

cluster: STABLE32 - rgmanager: Retry when config is out of sync
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=b2… Commit: b2012d8fe8b6a30f16091a8c96b5665e34892160 Parent: b4ca5bbb924ef9ef8e574a171e178f9440a28932 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Thu Aug 5 16:53:22 2010 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Wed Feb 22 10:31:44 2012 -0500 rgmanager: Retry when config is out of sync If you add a service to rgmanager v1 or v2 and that service fails to start on the first node but succeeds in its initial stop operation, there is a chance that the remote instance of rgmanager has not yet reread the configuration, causing the service to be placed into the 'recovering' state without further action. This patch causes the originator of the request to retry the operation. Later versions of rgmanager (ex STABLE3 branch and derivatives) are unlikely to have this problem since configuration updates are not polled, but rather delivered to clients. Update 22-Feb-2012: The above is incorrect, this was reproduced a rgmanager v3 installation. Resolves: rhbz#568126 Resolves: rhbz#796272 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/daemons/rg_state.c | 19 +++++++++++++++++++ 1 files changed, 19 insertions(+), 0 deletions(-) diff --git a/rgmanager/src/daemons/rg_state.c b/rgmanager/src/daemons/rg_state.c index 8b5dcaa..80b992a 100644 --- a/rgmanager/src/daemons/rg_state.c +++ b/rgmanager/src/daemons/rg_state.c @@ -1809,6 +1809,7 @@ handle_relocate_req(char *svcName, int orig_request, int preferred_target, rg_state_t svcStatus; int target = preferred_target, me = my_id(); int ret, x, request = orig_request; + int retries; get_rg_state_local(svcName, &svcStatus); if (svcStatus.rs_state == RG_STATE_DISABLED || @@ -1941,6 +1942,8 @@ handle_relocate_req(char *svcName, int orig_request, int preferred_target, if (target == me) goto exhausted; + retries = 0; +retry: ret = svc_start_remote(svcName, request, target); switch (ret) { case RG_ERUN: @@ -1950,6 +1953,22 @@ handle_relocate_req(char *svcName, int orig_request, int preferred_target, *new_owner = svcStatus.rs_owner; free_member_list(allowed_nodes); return 0; + case RG_ENOSERVICE: + /* + * Configuration update pending on remote node? Give it + * a few seconds to sync up. rhbz#568126 + * + * Configuration updates are synchronized in later releases + * of rgmanager; this should not be needed. + */ + if (retries++ < 4) { + sleep(3); + goto retry; + } + logt_print(LOG_WARNING, "Member #%d has a different " + "configuration than I do; trying next " + "member.", target); + /* Deliberate */ case RG_EDEPEND: case RG_EFAIL: /* Uh oh - we failed to relocate to this node.

1 0

cluster: RHEL6 - rgmanager: Fix dependency restart bug in CP mode
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=4f… Commit: 4fbd15844e11e3677e865779e57b5943ca6f3c99 Parent: 4758c35d2a9862bd7a8ac79d5c2239b6a79ca0bf Author: Adam Drew <adrew(a)redhat.com> AuthorDate: Fri Oct 7 12:31:16 2011 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Feb 21 13:37:11 2012 -0500 rgmanager: Fix dependency restart bug in CP mode If we: - had central_processing mode enabled, - had a service with 'depend' set on another service, - we attempted to restart the service with the dependency, and - the dependency was missing rgmanager would leave the state in 'recovering' This is the corrected patch from Adam. Resolves: rhbz#744824 Signed-off-by: Adam Drew <adrew(a)redhat.com> Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/resources/default_event_script.sl | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/rgmanager/src/resources/default_event_script.sl b/rgmanager/src/resources/default_event_script.sl index b14f466..c54c068 100644 --- a/rgmanager/src/resources/default_event_script.sl +++ b/rgmanager/src/resources/default_event_script.sl @@ -180,6 +180,7 @@ define move_or_start(service, node_list) ((event_type == EVENT_NODE) and (owner == node_id) and (node_state == 0))) { debug(service, " is not runnable; dependency not met"); + ()=service_stop(service); return ERR_DEPEND; } }

1 0

cluster: RHEL6 - rgmanager: Fix dependency issue related to ordering
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=47… Commit: 4758c35d2a9862bd7a8ac79d5c2239b6a79ca0bf Parent: 2543119ffd1ee25ff8e6f205f696f1f53a5292cf Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Thu Jan 5 10:43:12 2012 -0500 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Feb 21 13:34:21 2012 -0500 rgmanager: Fix dependency issue related to ordering If you ordered two services the opposite way in cluster.conf ('b' first, which depends on 'a'), then rgmanager would fail over 'b' despite the fact that 'a' was not running yet - this is because the state of 'a' was not set to 'stopped' - so we need to check for a node-death event in move_or_start(). Resolves: rhbz#743218 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/resources/default_event_script.sl | 4 +++- 1 files changed, 3 insertions(+), 1 deletions(-) diff --git a/rgmanager/src/resources/default_event_script.sl b/rgmanager/src/resources/default_event_script.sl index f3b35cd..b14f466 100644 --- a/rgmanager/src/resources/default_event_script.sl +++ b/rgmanager/src/resources/default_event_script.sl @@ -176,7 +176,9 @@ define move_or_start(service, node_list) depends = service_property(service, "depend"); if (depends != "") { (,,, owner, state) = service_status(depends); - if (owner < 0) { + if ((owner < 0) or + ((event_type == EVENT_NODE) and (owner == node_id) and + (node_state == 0))) { debug(service, " is not runnable; dependency not met"); return ERR_DEPEND; }

1 0

cluster: RHEL6 - rgmanager: Add -F to clusvcadm man page
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=fd… Commit: fdcec853307c5d9ce0517cb0088de2e970f76ae7 Parent: 4fbd15844e11e3677e865779e57b5943ca6f3c99 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Fri Jul 29 13:30:38 2011 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Feb 21 13:38:13 2012 -0500 rgmanager: Add -F to clusvcadm man page Resolves: rhbz#745226 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/man/clusvcadm.8 | 9 +++++++-- 1 files changed, 7 insertions(+), 2 deletions(-) diff --git a/rgmanager/man/clusvcadm.8 b/rgmanager/man/clusvcadm.8 index d5e09eb..055a0bf 100644 --- a/rgmanager/man/clusvcadm.8 +++ b/rgmanager/man/clusvcadm.8 @@ -8,7 +8,7 @@ clusvcadm \- Cluster User Service Administration Utility .B ] .B [\-e .I <service> -.B ] +.B [\-F] ] .B [\-l] .B [\-u] .B [\-S] @@ -52,10 +52,15 @@ on which the command is invoked. Stops and disables the user service named .I service -.IP "\-e <service>" +.IP "\-e <service> [\-F]" Enables and starts the user service named .I service + +Normally, rgmanager starts the service locally (i.e. on +the host where clusvcadm was run). However, if the \fB-F\fP +option is specified, rgmanager will use the assigned failover +domain rules as hints on where to start the service. .IP \-l Lock services in preparation for cluster shutdown. This should only be used if the administrator intends to perform a global, cluster

1 0

cluster: RHEL6 - rgmanager: Fix call to service_status()
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=23… Commit: 2353fc5a8b4f590d97e10e0c6cd2c91506eb853a Parent: 6563d1529e4929ad4094326da06ba1eb6b73c7c3 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Thu Jan 5 09:03:28 2012 -0500 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Feb 21 13:33:03 2012 -0500 rgmanager: Fix call to service_status() This patch simply fixes the calls to service_status(), the s_trans variable was being incorrectly set. Resolves: rhbz#743218 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/resources/default_event_script.sl | 7 +++++-- 1 files changed, 5 insertions(+), 2 deletions(-) diff --git a/rgmanager/src/resources/default_event_script.sl b/rgmanager/src/resources/default_event_script.sl index 763201d..f3b35cd 100644 --- a/rgmanager/src/resources/default_event_script.sl +++ b/rgmanager/src/resources/default_event_script.sl @@ -519,7 +519,7 @@ define default_service_event_handler() continue; } - (d_trans,,,, owner, state) = service_status(services[x], 1); + (d_trans,,,,, owner, state) = service_status(services[x], 1); if ((service_state == "started") and (owner < 0) and (state == "stopped")) { info("Dependency met; starting ", services[x]); @@ -537,9 +537,12 @@ define default_service_event_handler() % as above is running and the dependent service was % started at or after the service, then stopping it % will result in unwanted service outage. - (s_trans,,,, s_state) = service_status(service_name); + (s_trans,,,,,, s_state) = service_status(service_name, 1); if ((s_state == "started") and (state == "started") and (d_trans >= s_trans)) { + %debug("S:", service_name, " trans ", s_trans); + %debug("D:", services[x], " trans ", d_trans); + debug("Skipping ", services[x], "; restart not needed"); continue;

1 0

cluster: RHEL6 - rgmanager: Fix tiny memory leak in sl_service_status
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=25… Commit: 2543119ffd1ee25ff8e6f205f696f1f53a5292cf Parent: 2353fc5a8b4f590d97e10e0c6cd2c91506eb853a Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Thu Jan 5 10:12:05 2012 -0500 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Feb 21 13:33:45 2012 -0500 rgmanager: Fix tiny memory leak in sl_service_status Resolves: rhbz#743218 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/daemons/slang_event.c | 23 ++++++++++++++--------- 1 files changed, 14 insertions(+), 9 deletions(-) diff --git a/rgmanager/src/daemons/slang_event.c b/rgmanager/src/daemons/slang_event.c index c302de5..45110a3 100644 --- a/rgmanager/src/daemons/slang_event.c +++ b/rgmanager/src/daemons/slang_event.c @@ -209,7 +209,8 @@ get_service_state_internal(const char *svcName, rg_state_t *svcStatus) service_status(servicename) For extra information (flags, transition time) - (transition_time, flags, rte, restarts, owner, state) = + (transition_time, flags, rte, restarts, last_owner, + owner, state) = service_status(servicename, 1); */ static void @@ -273,7 +274,7 @@ sl_service_status(void) (char *)"%s: Failed to get status for %s", __FUNCTION__, svcName); - return; + goto out_free; } if (extra) { @@ -285,7 +286,7 @@ sl_service_status(void) (char *)"%s: Failed to push mtime %s", __FUNCTION__, svcName); - return; + goto out_free; } flags = (int)svcStatus.rs_flags; @@ -294,7 +295,7 @@ sl_service_status(void) (char *)"%s: Failed to push flags %s", __FUNCTION__, svcName); - return; + goto out_free; } } @@ -304,7 +305,7 @@ sl_service_status(void) (char *)"%s: Failed to push restarts_exceeded %s", __FUNCTION__, svcName); - return; + goto out_free; } if (SLang_push_integer(svcStatus.rs_restarts) < 0) { @@ -312,7 +313,7 @@ sl_service_status(void) (char *)"%s: Failed to push restarts for %s", __FUNCTION__, svcName); - return; + goto out_free; } if (SLang_push_integer(svcStatus.rs_last_owner) < 0) { @@ -320,7 +321,7 @@ sl_service_status(void) (char *)"%s: Failed to push last owner of %s", __FUNCTION__, svcName); - return; + goto out_free; } switch(svcStatus.rs_state) { @@ -338,7 +339,7 @@ sl_service_status(void) (char *)"%s: Failed to push owner of %s", __FUNCTION__, svcName); - return; + goto out_free; } if (svcStatus.rs_flags & RG_FLAG_FROZEN) { @@ -354,7 +355,7 @@ sl_service_status(void) (char *)"%s: Failed to duplicate state of %s", __FUNCTION__, svcName); - return; + goto out_free; } if (SLang_push_malloced_string(state_str) < 0) { @@ -364,6 +365,10 @@ sl_service_status(void) svcName); free(state_str); } + +out_free: + if (svcName) + free(svcName); }

1 0

cluster: RHEL6 - rgmanager: Fix clusvcadm message when run with -F
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=65… Commit: 6563d1529e4929ad4094326da06ba1eb6b73c7c3 Parent: 8168796b4f1d4e7435b2212a18d4cffbd86e1fa4 Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Mon Aug 1 16:31:57 2011 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Feb 21 13:27:22 2012 -0500 rgmanager: Fix clusvcadm message when run with -F The new_owner was not being correctly set when enabling a service with -F when run without central processing enabled. Resolves: rhbz#727326 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/daemons/rg_state.c | 1 + 1 files changed, 1 insertions(+), 0 deletions(-) diff --git a/rgmanager/src/daemons/rg_state.c b/rgmanager/src/daemons/rg_state.c index 5501b3f..23a4bec 100644 --- a/rgmanager/src/daemons/rg_state.c +++ b/rgmanager/src/daemons/rg_state.c @@ -2293,6 +2293,7 @@ handle_fd_start_req(char *svcName, int request, int *new_owner) switch(ret) { case RG_ESUCCESS: + *new_owner = target; ret = RG_ESUCCESS; goto out; case RG_ERUN:

1 0

cluster: RHEL6 - rgmanager: Resolve rare deadlock
by Lon Hohberger 22 Feb '12

22 Feb '12

Gitweb: http://git.fedorahosted.org/git/cluster.git?p=cluster.git;a=commitdiff;h=81… Commit: 8168796b4f1d4e7435b2212a18d4cffbd86e1fa4 Parent: 3624e3e0deed9eb3e6b4325a9ac3249dcb0a110a Author: Lon Hohberger <lhh(a)redhat.com> AuthorDate: Tue Sep 6 15:42:41 2011 -0400 Committer: Lon Hohberger <lhh(a)redhat.com> CommitterDate: Tue Feb 21 13:26:33 2012 -0500 rgmanager: Resolve rare deadlock In very rare cases, it was possible to fill up a pipe used for IPC between threads during reconfiguration, causing a deadlock. This patch resolves the issue. Resolves: rhbz#635152 Signed-off-by: Lon Hohberger <lhh(a)redhat.com> --- rgmanager/src/daemons/groups.c | 2 -- 1 files changed, 0 insertions(+), 2 deletions(-) diff --git a/rgmanager/src/daemons/groups.c b/rgmanager/src/daemons/groups.c index e9413cd..ee0fc81 100644 --- a/rgmanager/src/daemons/groups.c +++ b/rgmanager/src/daemons/groups.c @@ -1683,8 +1683,6 @@ init_resource_groups(int reconfigure, int do_init) free(val); } - /* Wait for any pending requests */ - rg_wait_threads(); /* Block operations that would break during configuration changes */ rg_clear_initialized(FL_CONFIG);

1 0

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

cluster-commits February 2012