Attachment 935688 Details for Bug 1072415 – proposed fix 1/2

[patch] proposed fix 1/2

0001-Add-RRP-support-in-adding-and-removing-nodes.patch (text/plain), 12.90 KB, created by Tomas Jelinek on 2014-09-09 13:10:03 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Tomas Jelinek

Created: 2014-09-09 13:10:03 UTC

Size: 12.90 KB

patch

obsolete

>From 8a67fd4e56bb5b9e40930294b8f80d717f037b5b Mon Sep 17 00:00:00 2001
>From: Tomas Jelinek <tojeline@redhat.com>
>Date: Mon, 8 Sep 2014 16:11:45 +0200
>Subject: [PATCH 1/2] Add RRP support in adding and removing nodes
>
>---
> pcs/cluster.py |  54 ++++++++++++++++++++++---------
> pcs/pcs.8      |   4 +--
> pcs/usage.py   |   6 ++--
> pcs/utils.py   | 100 +++++++++++++++++++++++++++++++++++++++------------------
> pcsd/remote.rb |   6 +++-
> 5 files changed, 119 insertions(+), 51 deletions(-)
>
>diff --git a/pcs/cluster.py b/pcs/cluster.py
>index 9c3366b..d174da6 100644
>--- a/pcs/cluster.py
>+++ b/pcs/cluster.py
>@@ -721,43 +721,67 @@ def cluster_node(argv):
>         sys.exit(1)
> 
>     node = argv[1]
>-    status,output = utils.checkAuthorization(node)
>+    if "," in node:
>+        node0 = node.split(",")[0]
>+        node1 = node.split(",")[1]
>+    else:
>+        node0 = node
>+        node1 = None
>+
>+    status,output = utils.checkAuthorization(node0)
>     if status == 2:
>-        utils.err("pcsd is not running on %s" % node)
>+        utils.err("pcsd is not running on %s" % node0)
>     elif status == 3:
>-        utils.err("%s is not yet authenticated (try pcs cluster auth %s)" % (node, node))
>+        utils.err(
>+            "%s is not yet authenticated (try pcs cluster auth %s)"
>+            % (node0, node0)
>+        )
> 
>     if add_node == True:
>+        if node1 is None and utils.need_ring1_address(utils.getCorosyncConf()):
>+            utils.err(
>+                "cluster is configured for RRP, "
>+                "you have to specify ring 1 address for the node"
>+            )
>+        elif (
>+            node1 is not None
>+            and
>+            not utils.need_ring1_address(utils.getCorosyncConf())
>+        ):
>+            utils.err(
>+                "cluster is not configured for RRP, "
>+                "you must not specify ring 1 address for the node"
>+            )
>         corosync_conf = None
>-        (canAdd, error) =  utils.canAddNodeToCluster(node)
>+        (canAdd, error) =  utils.canAddNodeToCluster(node0)
>         if not canAdd:
>-            utils.err("Unable to add '%s' to cluster: %s" % (node,error))
>+            utils.err("Unable to add '%s' to cluster: %s" % (node0, error))
> 
>         for my_node in utils.getNodesFromCorosyncConf():
>-            retval, output = utils.addLocalNode(my_node,node)
>+            retval, output = utils.addLocalNode(my_node, node0, node1)
>             if retval != 0:
>-                print >> sys.stderr, "Error: unable to add %s on %s - %s" % (node,my_node,output.strip())
>+                print >> sys.stderr, "Error: unable to add %s on %s - %s" % (node0, my_node, output.strip())
>             else:
>                 print "%s: Corosync updated" % my_node
>                 corosync_conf = output
>         if corosync_conf != None:
>-            utils.setCorosyncConfig(node, corosync_conf)
>+            utils.setCorosyncConfig(node0, corosync_conf)
>             if "--enable" in utils.pcs_options:
>-                utils.enableCluster(node)
>+                utils.enableCluster(node0)
>             if "--start" in utils.pcs_options:
>-                utils.startCluster(node)
>+                utils.startCluster(node0)
>         else:
>             utils.err("Unable to update any nodes")
>     else:
>         nodesRemoved = False
>         c_nodes = utils.getNodesFromCorosyncConf()
>-        destroy_cluster([node])
>+        destroy_cluster([node0])
>         for my_node in c_nodes:
>-            if my_node == node:
>+            if my_node == node0:
>                 continue
>-            retval, output = utils.removeLocalNode(my_node,node)
>+            retval, output = utils.removeLocalNode(my_node, node0)
>             if retval != 0:
>-                print >> sys.stderr, "Error: unable to remove %s on %s - %s" % (node,my_node,output.strip())
>+                print >> sys.stderr, "Error: unable to remove %s on %s - %s" % (node0,my_node,output.strip())
>             else:
>                 if output[0] == 0:
>                     print "%s: Corosync updated" % my_node
>@@ -767,7 +791,7 @@ def cluster_node(argv):
>         if nodesRemoved == False:
>             utils.err("Unable to update any nodes")
> 
>-        output, retval = utils.run(["crm_node", "--force","-R", node])
>+        output, retval = utils.run(["crm_node", "--force", "-R", node0])
> 
> def cluster_localnode(argv):
>     if len(argv) != 2:
>diff --git a/pcs/pcs.8 b/pcs/pcs.8
>index ac09816..1991cf5 100644
>--- a/pcs/pcs.8
>+++ b/pcs/pcs.8
>@@ -235,8 +235,8 @@ Rollback the CIB to specified revision.  Use the cib-revisions command to get a
> edit
> Edit the cib in the editor specified by the $EDITOR environment variable and push out any changes upon saving
> .TP
>-node add <node> [\fB\-\-start\fR] [\fB\-\-enable\fR]
>-Add the node to corosync.conf and corosync on all nodes in the cluster and sync the new corosync.conf to the new node.  If \fB\-\-start\fR is specified also start corosync/pacemaker on the new node, if \fB\-\-enable\fR is specified enable corosync/pacemaker on new node
>+node add <node[,node\-altaddr]> [\fB\-\-start\fR] [\fB\-\-enable\fR]
>+Add the node to corosync.conf and corosync on all nodes in the cluster and sync the new corosync.conf to the new node.  If \fB\-\-start\fR is specified also start corosync/pacemaker on the new node, if \fB\-\-enable\fR is specified enable corosync/pacemaker on new node.  When using Redundant Ring Protocol (RRP) with udpu transport, specify the ring 0 address first followed by a ',' and then the ring 1 address.
> .TP
> node remove <node>
> Shutdown specified node and remove it from pacemaker and corosync on all other nodes in the cluster
>diff --git a/pcs/usage.py b/pcs/usage.py
>index 2e8e62a..e3ccace 100644
>--- a/pcs/usage.py
>+++ b/pcs/usage.py
>@@ -582,11 +582,13 @@ Commands:
>         Edit the cib in the editor specified by the $EDITOR environment
>         variable and push out any changes upon saving
> 
>-    node add <node> [--start] [--enable]
>+    node add <node[,node-altaddr]> [--start] [--enable]
>         Add the node to corosync.conf and corosync on all nodes in the cluster
>         and sync the new corosync.conf to the new node.  If --start is specified
>         also start corosync/pacemaker on the new node, if --enable is specified
>-        enable corosync/pacemaker on new node
>+        enable corosync/pacemaker on new node.
>+        When using Redundant Ring Protocol (RRP) with udpu transport, specify
>+        the ring 0 address first followed by a ',' and then the ring 1 address.
> 
>     node remove <node>
>         Shutdown specified node and remove it from pacemaker and corosync on
>diff --git a/pcs/utils.py b/pcs/utils.py
>index d85aba1..7df43f8 100644
>--- a/pcs/utils.py
>+++ b/pcs/utils.py
>@@ -234,8 +234,11 @@ def canAddNodeToCluster(node):
> 
>     return (False,"error checking node availability")
> 
>-def addLocalNode(node,node_to_add):
>-    data = urllib.urlencode({'new_nodename':node_to_add})
>+def addLocalNode(node, node_to_add, ring1_addr=None):
>+    options = {'new_nodename': node_to_add}
>+    if ring1_addr:
>+        options['new_ring1addr'] = ring1_addr
>+    data = urllib.urlencode(options)
>     retval, output = sendHTTPRequest(node, 'remote/add_node', data, False, False)
>     if retval == 0:
>         try:
>@@ -418,15 +421,21 @@ def getCorosyncActiveNodes():
> # Add node specified to corosync.conf and reload corosync.conf (if running)
> def addNodeToCorosync(node):
> # Before adding, make sure node isn't already in corosync.conf
>+    if "," in node:
>+        node0 = node.split(",")[0]
>+        node1 = node.split(",")[1]
>+    else:
>+        node0 = node
>+        node1 = None
>     used_node_ids = []
>     num_nodes_in_conf = 0
>     for c_node in getNodesFromCorosyncConf():
>-        if c_node == node:
>+        if (c_node == node0) or (c_node == node1):
>             err("node already exists in corosync.conf")
>         num_nodes_in_conf = num_nodes_in_conf + 1
>     if "--corosync_conf" not in pcs_options:
>         for c_node in getCorosyncActiveNodes():
>-            if c_node == node:
>+            if (c_node == node0) or (c_node == node1):
>                 err("Node already exists in running corosync")
>     corosync_conf = getCorosyncConf()
>     new_nodeid = getNextNodeID(corosync_conf)
>@@ -446,11 +455,11 @@ def addNodeToCorosync(node):
>             count += 1
>         new_corosync_conf = corosync_conf[:count]
>         new_corosync_conf += "  node {\n"
>-        if "," in node:
>-            new_corosync_conf += "        ring0_addr: %s\n" % (node.split(",")[0])
>-            new_corosync_conf += "        ring1_addr: %s\n" % (node.split(",")[1])
>+        if node1 is not None:
>+            new_corosync_conf += "        ring0_addr: %s\n" % (node0)
>+            new_corosync_conf += "        ring1_addr: %s\n" % (node1)
>         else:
>-            new_corosync_conf += "        ring0_addr: %s\n" % (node)
>+            new_corosync_conf += "        ring0_addr: %s\n" % (node0)
>         new_corosync_conf += "        nodeid: %d\n" % (new_nodeid)
>         new_corosync_conf += "       }\n"
>         new_corosync_conf += corosync_conf[count:]
>@@ -490,19 +499,16 @@ def addNodeToClusterConf(node):
> 
> # TODO: Need to make this smarter about parsing files not generated by pcs
> def removeNodeFromCorosync(node):
>-    rrp = False
>-    if "," in node:
>-        rrp = True
>-
>     removed_node = False
>     node_found = False
>     num_nodes_in_conf = 0
> 
>-    if rrp:
>+    if "," in node:
>         node0 = node.split(",")[0]
>         node1 = node.split(",")[1]
>     else:
>         node0 = node
>+        node1 = None
> 
>     for c_node in getNodesFromCorosyncConf():
>         if c_node == node0:
>@@ -512,25 +518,37 @@ def removeNodeFromCorosync(node):
>     if not node_found:
>         return False
> 
>-    corosync_conf = getCorosyncConf().split("\n")
>-    for x in range(len(corosync_conf)):
>-        if corosync_conf[x].find("node {") != -1:
>-
>-            match = False
>-            if (rrp and corosync_conf[x+1].find("ring0_addr: "+node0 ) != -1 and corosync_conf[x+2].find("ring1_addr: "+node1 ) != -1) or (not rrp and corosync_conf[x+1].find("ring0_addr: "+node0) != -1):
>-                match = True
>-
>-            if match:
>-                if rrp:
>-                    new_corosync_conf = "\n".join(corosync_conf[0:x] + corosync_conf[x+5:])
>-                else:
>-                    new_corosync_conf = "\n".join(corosync_conf[0:x] + corosync_conf[x+4:])
>-
>-                if num_nodes_in_conf == 3:
>-                    new_corosync_conf = addQuorumOption(new_corosync_conf,("two_node","1"))
>-                setCorosyncConf(new_corosync_conf)
>-                reloadCorosync()
>+    new_corosync_conf_lines = []
>+    in_node = False
>+    node_match = False
>+    node_buffer = []
>+    for line in getCorosyncConf().split("\n"):
>+        if in_node:
>+            node_buffer.append(line)
>+            if (
>+                ("ring0_addr: " + node0 in line)
>+                or
>+                (node1 is not None and "ring0_addr: " + node1 in line)
>+            ):
>+                node_match = True
>                 removed_node = True
>+            if "}" in line:
>+                if not node_match:
>+                    new_corosync_conf_lines.extend(node_buffer)
>+                node_buffer = []
>+                node_match = False
>+        elif "node {" in line:
>+            node_buffer.append(line)
>+            in_node = True
>+        else:
>+            new_corosync_conf_lines.append(line)
>+    new_corosync_conf = "\n".join(new_corosync_conf_lines) + "\n"
>+
>+    if removed_node:
>+        if num_nodes_in_conf == 3:
>+            new_corosync_conf = addQuorumOption(new_corosync_conf,("two_node","1"))
>+        setCorosyncConf(new_corosync_conf)
>+        reloadCorosync()
> 
>     return removed_node
> 
>@@ -624,6 +642,26 @@ def getNextNodeID(corosync_conf):
> 
>     return highest + 1
> 
>+def need_ring1_address(corosync_conf):
>+    line_list = corosync_conf.split("\n")
>+    in_totem = False
>+    udpu_transport = False
>+    rrp = False
>+    for line in line_list:
>+        line = line.strip()
>+        if in_totem:
>+            if ":" in line:
>+                name, value = map(lambda x: x.strip(), line.split(":"))
>+                if name == "transport" and value == "udpu":
>+                    udpu_transport = True
>+                if name == "rrp_mode" and value in ["active", "passive"]:
>+                    rrp = True
>+            if "}" in line:
>+                in_totem = False
>+        if line.startswith("totem {"):
>+            in_totem = True
>+    return udpu_transport and rrp
>+
> # Restore default behavior before starting subprocesses
> def subprocess_setup():
>     signal.signal(signal.SIGPIPE, signal.SIG_DFL)
>diff --git a/pcsd/remote.rb b/pcsd/remote.rb
>index 8fcf981..3e7ef01 100644
>--- a/pcsd/remote.rb
>+++ b/pcsd/remote.rb
>@@ -333,7 +333,11 @@ def remote_add_node(params,all = false)
>   end
> 
>   if params[:new_nodename] != nil
>-    retval, output =  add_node(params[:new_nodename],all,auto_start)
>+    node = params[:new_nodename]
>+    if params[:new_ring1addr] != nil
>+      node += ',' + params[:new_ring1addr]
>+    end
>+    retval, output = add_node(node, all, auto_start)
>   end
> 
>   if retval == 0
>-- 
>1.9.1
>

Actions: View | Diff

Attachments on bug 1072415: 935688 | 935689 | 940740