Attachment 148657 Details for Bug 229650 – Patch that will not handle the special flag discussed in comments #1 and #2.

[patch] Patch that will not handle the special flag discussed in comments #1 and #2.

rgmanager-cvs-recover_only_failed_resources-v01.patch (text/plain), 5.48 KB, created by Simone Gotti on 2007-02-23 10:06:21 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Simone Gotti

Created: 2007-02-23 10:06:21 UTC

Size: 5.48 KB

patch

obsolete

>Index: src/daemons/restree.c
>===================================================================
>RCS file: /cvs/cluster/cluster/rgmanager/src/daemons/restree.c,v
>retrieving revision 1.25
>diff -u -p -r1.25 restree.c
>--- src/daemons/restree.c	22 Feb 2007 16:37:28 -0000	1.25
>+++ src/daemons/restree.c	23 Feb 2007 09:53:42 -0000
>@@ -1010,7 +1010,7 @@ _res_op(resource_node_t **tree, resource
> 			}
> 		}
> 
>-		/* Start starts before children */
>+		/* Start/status starts before children */
> 		if (me && (op == RS_START)) {
> 			node->rn_flags &= ~RF_NEEDSTART;
> 
>@@ -1030,8 +1030,15 @@ _res_op(resource_node_t **tree, resource
> 		} else if (me && (op == RS_STATUS)) {
> 			/* Check status before children*/
> 			rv = do_status(node);
>-			if (rv != 0)
>+			if (rv != 0) {
>+				/* This resource needs a condstop and a
>+				 * condstart as it returned and error during
>+				 * its status check
>+				 */
>+				node->rn_flags |= RF_NEEDSTOP;
>+				node->rn_flags |= RF_NEEDSTART;
> 				return rv;
>+			}
> 		}
> 
> 
>Index: src/daemons/rg_state.c
>===================================================================
>RCS file: /cvs/cluster/cluster/rgmanager/src/daemons/rg_state.c,v
>retrieving revision 1.28
>diff -u -p -r1.28 rg_state.c
>--- src/daemons/rg_state.c	20 Feb 2007 19:54:42 -0000	1.28
>+++ src/daemons/rg_state.c	23 Feb 2007 09:53:44 -0000
>@@ -713,7 +713,7 @@ svc_advise_start(rg_state_t *svcStatus, 
>  * @return		FAIL, 0
>  */
> int
>-svc_start(char *svcName, int req)
>+_svc_start(char *svcName, int req, int condstart)
> {
> 	struct dlm_lksb lockp;
> 	int ret;
>@@ -770,7 +770,11 @@ svc_start(char *svcName, int req)
> 	
> 	rg_unlock(&lockp);
> 
>-	ret = group_op(svcName, RG_START);
>+	if (condstart)
>+		ret = group_op(svcName, RG_CONDSTART);
>+	else
>+		ret = group_op(svcName, RG_START);
>+
> 	ret = !!ret; /* Either it worked or it didn't.  Ignore all the
> 			cute values scripts might return */
> 
>@@ -804,6 +808,18 @@ svc_start(char *svcName, int req)
> 	return ret;
> }
> 
>+int
>+svc_start(char *svcName, int req)
>+{
>+	return _svc_start(svcName, req, 0);
>+}
>+
>+int
>+svc_condstart(char *svcName, int req)
>+{
>+	return _svc_start(svcName, req, 1);
>+}
>+
> 
> /**
>  * Migrate a service to another node.
>@@ -950,7 +966,7 @@ svc_status(char *svcName)
>  * @return		FAIL, 0
>  */
> static int
>-_svc_stop(char *svcName, int req, int recover, uint32_t newstate)
>+_svc_stop(char *svcName, int req, int recover, int condstop, uint32_t newstate)
> {
> 	struct dlm_lksb lockp;
> 	rg_state_t svcStatus;
>@@ -1014,7 +1030,11 @@ _svc_stop(char *svcName, int req, int re
> 	}
> 	rg_unlock(&lockp);
> 
>-	ret = group_op(svcName, RG_STOP);
>+	if (condstop)
>+		/* To recover a resource stop only from the failed childs and not the whole RG */
>+		ret = group_op(svcName, RG_CONDSTOP);
>+	else
>+		ret = group_op(svcName, RG_STOP);
> 
> 	if (old_state == RG_STATE_FAILED && newstate == RG_STATE_DISABLED) {
> 		if (ret)
>@@ -1098,14 +1118,21 @@ _svc_stop_finish(char *svcName, int fail
> int
> svc_disable(char *svcName)
> {
>-	return _svc_stop(svcName, RG_DISABLE, 0, RG_STATE_DISABLED);
>+	return _svc_stop(svcName, RG_DISABLE, 0, 0, RG_STATE_DISABLED);
> }
> 
> 
> int
> svc_stop(char *svcName, int req)
> {
>-	return _svc_stop(svcName, req, (req == RG_STOP_RECOVER),
>+	return _svc_stop(svcName, req, (req == RG_STOP_RECOVER), 0,
>+			 RG_STATE_STOPPED);
>+}
>+
>+int
>+svc_condstop(char *svcName, int req)
>+{
>+	return _svc_stop(svcName, req, (req == RG_STOP_RECOVER), 1,
> 			 RG_STATE_STOPPED);
> }
> 
>@@ -1295,7 +1322,7 @@ handle_relocate_req(char *svcName, int r
> 	 * Stop the service - if we haven't already done so.
> 	 */
> 	if (request != RG_START_RECOVER) {
>-		ret = _svc_stop(svcName, request, 0, RG_STATE_STOPPED);
>+		ret = _svc_stop(svcName, request, 0, 0, RG_STATE_STOPPED);
> 		if (ret == RG_EFAIL) {
> 			svc_fail(svcName);
> 			return RG_EFAIL;
>@@ -1486,7 +1513,10 @@ handle_start_req(char *svcName, int req,
> 	 * This is a 'root' start request.  We need to clear out our failure
> 	 * mask here - so that we can try all nodes if necessary.
> 	 */
>-	ret = svc_start(svcName, req);
>+	if (req == RG_START_RECOVER) 
>+		ret = svc_condstart(svcName, req);
>+	else
>+		ret = svc_start(svcName, req);
> 
> 	/* 
> 	   If services are locked, return the error 
>@@ -1603,9 +1633,19 @@ int
> handle_recover_req(char *svcName, int *new_owner)
> {
> 	char policy[20];
>+	int ret;
> 
> 	get_recovery_policy(svcName, policy, sizeof(policy));
> 
>+	if ((!strcasecmp(policy, "disable")) || (!strcasecmp(policy, "relocate"))) {
>+		/* Try to stop the service */
>+		if ((ret = svc_stop(svcName, RG_STOP_RECOVER)) != 0)
>+			return ret;
>+	} else {
>+		if ((ret = svc_condstop(svcName, RG_STOP_RECOVER)) != 0)
>+			return ret;
>+	}
>+		
> 	if (!strcasecmp(policy, "disable")) {
> 		return svc_disable(svcName);
> 	} else if (!strcasecmp(policy, "relocate")) {
>Index: src/daemons/rg_thread.c
>===================================================================
>RCS file: /cvs/cluster/cluster/rgmanager/src/daemons/rg_thread.c,v
>retrieving revision 1.17
>diff -u -p -r1.17 rg_thread.c
>--- src/daemons/rg_thread.c	14 Dec 2006 20:11:04 -0000	1.17
>+++ src/daemons/rg_thread.c	23 Feb 2007 09:53:44 -0000
>@@ -403,18 +403,15 @@ resgroup_thread_main(void *arg)
> 			   resource groups we don't own */
> 			error = svc_status(myname);
> 
>-			/* Recover dead service */
> 			if (error == 0) {
> 				ret = RG_SUCCESS;
> 				break;
> 			}
> 
>-			error = svc_stop(myname, RG_STOP_RECOVER);
>-			if (error == 0) {
>-				error = handle_recover_req(myname, &newowner);
>-				if (error == 0)
>-					ret = RG_SUCCESS;
>-			}
>+			/* Recover dead service */
>+			error = handle_recover_req(myname, &newowner);
>+			if (error == 0)
>+				ret = RG_SUCCESS;
> 
> 			break;
>

Actions: View | Diff

Attachments on bug 229650: 148657 | 155385 | 155386 | 155387 | 155388 | 155389 | 155835 | 156923 | 181341 | 205561 | 205861