From a0279514893dbfa063b36791a5a0f82b3759be52 Mon Sep 17 00:00:00 2001
From: johann <johann>
Date: Fri, 31 Aug 2007 12:48:09 +0000
Subject: [PATCH] Branch HEAD b=12459 i=adilger,tianzy i=scjody
Severity : normal
Bugzilla : 12459
Description: Client eviction due to failover config
Details : after a connection loss, the lustre client should attempt to
reconnect to the last active server first before trying the
other potential connections.
---
lustre/ChangeLog | 7 ++++++
lustre/ptlrpc/import.c | 2 +-
lustre/tests/conf-sanity.sh | 46 +++++++++++++++++++++++++++++++++++++
3 files changed, 54 insertions(+), 1 deletion(-)
diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 59e69a5003..bacd497099 100644
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -201,6 +201,13 @@ Bugzilla : 11802
Description: lustre support for RHEL5
Details : Add support for RHEL5.
+Severity : normal
+Bugzilla : 12459
+Description: Client eviction due to failover config
+Details : after a connection loss, the lustre client should attempt to
+ reconnect to the last active server first before trying the
+ other potential connections.
+
--------------------------------------------------------------------------------
2007-08-10 Cluster File Systems, Inc. <info@clusterfs.com>
diff --git a/lustre/ptlrpc/import.c b/lustre/ptlrpc/import.c
index 07785302d3..0c7eff2392 100644
--- a/lustre/ptlrpc/import.c
+++ b/lustre/ptlrpc/import.c
@@ -290,7 +290,7 @@ static int import_select_connection(struct obd_import *imp)
cfs_time_current_64())) {
/* If we have never tried this connection since the
the last successful attempt, go with this one */
- if (cfs_time_before_64(conn->oic_last_attempt,
+ if (cfs_time_beforeq_64(conn->oic_last_attempt,
imp->imp_last_success_conn)) {
imp_conn = conn;
break;
diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh
index 46f36ed66e..a61e502e48 100644
--- a/lustre/tests/conf-sanity.sh
+++ b/lustre/tests/conf-sanity.sh
@@ -1193,6 +1193,52 @@ test_33() { # bug 12333
}
run_test 33 "Mount ost with a large index number"
+test_35() { # bug 12459
+ setup
+
+ DBG_SAVE="`sysctl -n lnet.debug`"
+ sysctl -w lnet.debug="ha"
+
+ log "Set up a fake failnode for the MDS"
+ FAKENID="127.0.0.2"
+ $LCTL conf_param ${FSNAME}-MDT0000.failover.node=$FAKENID || return 4
+
+ log "Wait for RECONNECT_INTERVAL seconds (10s)"
+ sleep 10
+
+ MSG="conf-sanity.sh test_33 `date +%F%kh%Mm%Ss`"
+ $LCTL clear
+ log "$MSG"
+ log "Stopping the MDT:"
+ stop_mds || return 5
+
+ df $MOUNT > /dev/null 2>&1 &
+ DFPID=$!
+ log "Restarting the MDT:"
+ start_mds || return 6
+ log "Wait for df ($DFPID) ... "
+ wait $DFPID
+ log "done"
+ sysctl -w lnet.debug="$DBG_SAVE"
+
+ # retrieve from the log the first server that the client tried to
+ # contact after the connection loss
+ $LCTL dk $TMP/lustre-log-$TESTNAME.log
+ NEXTCONN=`awk "/${MSG}/ {start = 1;}
+ /import_select_connection.*${FSNAME}-MDT0000-mdc.* using connection/ {
+ if (start) {
+ if (\\\$NF ~ /$FAKENID/)
+ print \\\$NF;
+ else
+ print 0;
+ exit;
+ }
+ }" $TMP/lustre-log-$TESTNAME.log`
+ [ "$NEXTCONN" != "0" ] && log "The client didn't try to reconnect to the last active server (tried ${NEXTCONN} instead)" && return 7
+ cleanup
+}
+run_test 35 "Reconnect to the last active server first"
+
umount_client $MOUNT
cleanup_nocli
cleanup_krb5_env
--
GitLab