Attachment 6609 Details for Bug 8080 – ctdb-client-connect-retry.patch

[patch] ctdb-client-connect-retry.patch

ctdb-client-connect-retry.patch (text/plain), 2.45 KB, created by David Disseldorp on 2011-06-22 16:46:05 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: David Disseldorp

Created: 2011-06-22 16:46:05 UTC

Size: 2.45 KB

patch

obsolete

>From c9ce2ea476c88cb59d5174288edd835fe905f719 Mon Sep 17 00:00:00 2001
>From: David Disseldorp <ddiss@suse.de>
>Date: Tue, 5 Apr 2011 13:26:29 +0200
>Subject: [PATCH 1/2] client: handle transient connection errors
>
>Client connections to the ctdbd unix domain socket may fail
>intermittently while the server is under heavy load. This change
>introduces a client connect retry loop.
>
>During failure the client will retry for a maximum of 64 seconds, the
>ctdb --timelimit option can be used to cap client runtime.
>---
> client/ctdb_client.c |   35 ++++++++++++++++++++++++++++++-----
> 1 files changed, 30 insertions(+), 5 deletions(-)
>
>diff --git a/client/ctdb_client.c b/client/ctdb_client.c
>index 2d3c176..3edcd1a 100644
>--- a/client/ctdb_client.c
>+++ b/client/ctdb_client.c
>@@ -246,16 +246,41 @@ done:
> }
> 
> /*
>-  connect to a unix domain socket
>+  connect with exponential backoff, thanks Stevens
> */
>-int ctdb_socket_connect(struct ctdb_context *ctdb)
>+#define CONNECT_MAXSLEEP 64
>+static int ctdb_connect_retry(struct ctdb_context *ctdb)
> {
> 	struct sockaddr_un addr;
>+	int secs;
>+	int ret = 0;
> 
> 	memset(&addr, 0, sizeof(addr));
> 	addr.sun_family = AF_UNIX;
> 	strncpy(addr.sun_path, ctdb->daemon.name, sizeof(addr.sun_path));
> 
>+	for (secs = 1; secs <= CONNECT_MAXSLEEP; secs *= 2) {
>+		ret = connect(ctdb->daemon.sd, (struct sockaddr *)&addr,
>+			      sizeof(addr));
>+		if ((ret == 0) || (errno != EAGAIN)) {
>+			break;
>+		}
>+
>+		if (secs <= (CONNECT_MAXSLEEP / 2)) {
>+			DEBUG(DEBUG_ERR,("connect failed: %s, retry in %d second(s)\n",
>+					 strerror(errno), secs));
>+			sleep(secs);
>+		}
>+	}
>+
>+	return ret;
>+}
>+
>+/*
>+  connect to a unix domain socket
>+*/
>+int ctdb_socket_connect(struct ctdb_context *ctdb)
>+{
> 	ctdb->daemon.sd = socket(AF_UNIX, SOCK_STREAM, 0);
> 	if (ctdb->daemon.sd == -1) {
> 		DEBUG(DEBUG_ERR,(__location__ " Failed to open client socket. Errno:%s(%d)\n", strerror(errno), errno));
>@@ -264,11 +289,11 @@ int ctdb_socket_connect(struct ctdb_context *ctdb)
> 
> 	set_nonblocking(ctdb->daemon.sd);
> 	set_close_on_exec(ctdb->daemon.sd);
>-	
>-	if (connect(ctdb->daemon.sd, (struct sockaddr *)&addr, sizeof(addr)) == -1) {
>+
>+	if (ctdb_connect_retry(ctdb) == -1) {
>+		DEBUG(DEBUG_ERR,(__location__ " Failed to connect client socket to daemon. Errno:%s(%d)\n", strerror(errno), errno));
> 		close(ctdb->daemon.sd);
> 		ctdb->daemon.sd = -1;
>-		DEBUG(DEBUG_ERR,(__location__ " Failed to connect client socket to daemon. Errno:%s(%d)\n", strerror(errno), errno));
> 		return -1;
> 	}
> 
>-- 
>1.7.1
>

Actions: View

Attachments on bug 8080: 6401 | 6609