LU-13688 hsm: handle in-tree executed copytools correctly

The Lustre test suite and HSM copytools can be invoked from either
within /usr/lib{,64}/ if they have been installed from source or from
packages, or from within the Lustre source tree, usually for
development purposes; in the latter case, the copytool process name is
prepended with an "lt-", due to being invoked via a libtool wrapper
script. The Lustre test framework relies on "libtool execute" to
distinguish between these two cases, parse the command parameters and
pass the correct process name as a parameter to utilities such as
pgrep(1), pkill(1), ps(1) and killall(1). Unfortunately, this doesn't
seem to work unless the libtool script for the copytool and the test
framework test file are in the same directory; e.g. this doesn't work
for lhsmtool_posix as its libtool script is in lustre/utils/, but the
Lustre test suite is in lustre/tests/, which doesn't allow the
"libtool execute" parsing and parameter replacing to succeed.

Fix this by determining the process name of the executed copytool
based on whether it was invoked from within the source tree or not and
using it in commands that either search for copytool processes or send
them signals by process name.

Signed-off-by: Nikitas Angelinas <nikitas.angelinas@hpe.com>
Cray-bug-id: LUS-8931
Change-Id: Ief7b224b793401b1a24bf9780d1df6e029f5c0d7
Reviewed-on: https://review.whamcloud.com/38962
Tested-by: jenkins <devops@whamcloud.com>
Reviewed-by: James Simmons <jsimmons@infradead.org>
Reviewed-by: Ben Evans <beevans@whamcloud.com>
Reviewed-by: nathan r <nrutman@gmail.com>
Tested-by: Maloo <maloo@whamcloud.com>
Reviewed-by: Oleg Drokin <green@whamcloud.com>

diff --git a/lustre/tests/sanity-hsm.sh b/lustre/tests/sanity-hsm.sh
index 15f9551..01f7e0a 100755 (executable)
--- a/lustre/tests/sanity-hsm.sh
+++ b/lustre/tests/sanity-hsm.sh
@@ -84,7 +84,7 @@ CLIENT2=${CLIENT2:-$CLIENT1}
 
 search_copytools() {
        local hosts=${1:-$(facet_active_host $SINGLEAGT)}
-       do_nodesv $hosts "libtool execute pgrep -x $HSMTOOL"
+       do_nodesv $hosts "pgrep -x $HSMTOOL_PKILL"
 }
 
 wait_copytools() {
@@ -170,8 +170,8 @@ copytool_suspend() {
        local agents=${1:-$(facet_active_host $SINGLEAGT)}
 
        stack_trap \
-               "do_nodesv $agents libtool execute pkill -CONT -x '$HSMTOOL' || true" EXIT
-       do_nodesv $agents "libtool execute pkill -STOP -x $HSMTOOL" || return 0
+               "do_nodesv $agents pkill -CONT -x '$HSMTOOL_PKILL' || true" EXIT
+       do_nodesv $agents "pkill -STOP -x $HSMTOOL_PKILL" || return 0
        echo "Copytool is suspended on $agents"
 }
 
@@ -489,7 +489,7 @@ get_agent_uuid() {
 
        # Lustre mount-point is mandatory and last parameter on
        # copytool cmd-line.
-       local mntpnt=$(do_rpc_nodes $agent libtool execute ps -C $HSMTOOL -o args= |
+       local mntpnt=$(do_rpc_nodes $agent ps -C $HSMTOOL_KILL -o args= |
                       awk '{print $NF}')
        [ -n "$mntpnt" ] || error "Found no Agent or with no mount-point "\
                                  "parameter"

diff --git a/lustre/tests/test-framework.sh b/lustre/tests/test-framework.sh
index fd84e2d..65aeb10 100755 (executable)
--- a/lustre/tests/test-framework.sh
+++ b/lustre/tests/test-framework.sh
@@ -10034,6 +10034,15 @@ init_agt_vars() {
        export HSMTOOL_EVENT_FIFO=${HSMTOOL_EVENT_FIFO:=""}
        export HSMTOOL_TESTDIR
 
+       # Copytools invoked from within the source tree are executed through a
+       # libtool script and have a different process name
+       from_build_tree && LIBTOOL_PREFIX=lt-
+
+       export HSMTOOL_KILL=${LIBTOOL_PREFIX}${HSMTOOL}
+
+       # pgrep(1) and pkill(1) limit process name matching to 15 characters
+       export HSMTOOL_PKILL=${HSMTOOL_KILL:0:15}
+
        HSM_ARCHIVE_NUMBER=2
 
        # The test only support up to 10 MDTs
@@ -10069,7 +10078,7 @@ get_mdt_devices() {
 copytool_continue() {
        local agents=${1:-$(facet_active_host $SINGLEAGT)}
 
-       do_nodesv $agents "libtool execute pkill -CONT -x $HSMTOOL" || return 0
+       do_nodesv $agents "pkill -CONT -x $HSMTOOL_PKILL" || return 0
        echo "Copytool is continued on $agents"
 }
 
@@ -10077,7 +10086,7 @@ kill_copytools() {
        local hosts=${1:-$(facet_active_host $SINGLEAGT)}
 
        echo "Killing existing copytools on $hosts"
-       do_nodesv $hosts "libtool execute killall -q $HSMTOOL" || true
+       do_nodesv $hosts "killall -q $HSMTOOL_KILL" || true
        copytool_continue "$hosts"
 }
 
@@ -10133,7 +10142,7 @@ __lhsmtool_setup()
        cmd+=" \"$mountpoint\""
 
        echo "Starting copytool $facet on $(facet_host $facet)"
-       stack_trap "do_facet $facet libtool execute pkill -x '$HSMTOOL' || true" EXIT
+       stack_trap "do_facet $facet pkill -x '$HSMTOOL_PKILL' || true" EXIT
        do_facet $facet "$cmd < /dev/null > \"$(copytool_logfile $facet)\" 2>&1"
 }