* Cleanup

* Catch up to build ID directory changes

* Support a meta-hostname of 'all' for setting up all clients at once.
  This is better than the old way of running one copy of the script
  for each client by hand, since it is easier and involves less
  duplicated work.

* We copy in the per-build ports, src, and bindist .tbz files and .md5
  checksums, as well as refreshing the build scripts and
  bindist-$(hostname).tar customization tarball.

* The -force switch forces copying of files and re-extraction of the
  tarballs on the client.  This is necessary in order to propagate
  local changes to the tarballs after the initial client setup
  (e.g. if you need to change a file in the ports tree, it must be
  recompressed, redistributed, and re-extracted on the client).

* The -queue switch will poll the client's job queue after completion
  of the setup.  This is racy and should only be used when the machine
  is not currently accepting jobs.

* For cleaning up a build the 'build cleanup' command should now be
  used instead.  It calls back into this command but also allows full
  clenaup of build-local files on the client.

TODO: "all" setups are hard on the server since they may spawn dozens
of rsyncs at once.  A better solution would be to have a worker pool
of setup tasks to limit the maximum load.

1 files changed, 153 insertions, 28 deletions

diff --git a/Tools/portbuild/scripts/dosetupnode b/Tools/portbuild/scripts/dosetupnode
index 9dc5edfebae3..33a6044a627f 100755
--- a/Tools/portbuild/scripts/dosetupnode
+++ b/Tools/portbuild/scripts/dosetupnode
@@ -1,45 +1,170 @@
 #!/bin/sh
+#
+# XXX Use a worker pool that only runs N setups at once to avoid
+# raping the server.  Hard to do in shell?
+
+# -norsync|-nocopy : Don't copy files, just clean up builds
+#
+# -queue : update queue once we finish setting up
+#
+# -force : force file copying/extraction even if it appears it is
+#          up-to-date
+#
+# NB: branch or buildid might be "-" to specify only to set up the
+# scripts/ and ${arch}/ directories (e.g. after client reboot)
 
 # configurable variables
 pb=/var/portbuild
 arch=$1
 branch=$2
-node=$3
-shift 3
+buildid=$3
+nodelist=$4
+shift 4
 
-norsync=0
-if [ "$1" = "-norsync" ]; then
-  norsync=1
+if [ -f ${pb}/${arch}/portbuild.conf ]; then
+    . ${pb}/${arch}/portbuild.conf
+else
+    echo "Invalid arch ${arch}"
+    exit 1
 fi
+. ${pb}/scripts/buildenv
 
-. ${pb}/${arch}/portbuild.conf
+# Check for non-fatal rsync errors
+checkerror() {
+    error=$?
 
-. ${pb}/scripts/buildenv
+    case $error in
+	0)
+	    return 0
+	    ;;
+	23)
+	    echo "Continuing..."	 
+	    return 0
+	    ;;
+	*)
+	    echo "Aborting..."
+	    return 1
+	    ;;
+    esac
+}
 
-echo "setting up of $node started at $(date)"
+setup() {
+    node=$1
 
-# Read in client config and decide if the node is disconnected or not (i.e. mounts via NFS)
-disconnected=0
-. ${pb}/${arch}/portbuild.${node}
+    echo "setting up of $node started at $(date)"
 
-if [ "${norsync}" = 1 ]; then
-  md5=0
-  nocopy=-nocopy
-else
-  nocopy=
-  md5=$(/sbin/md5 ${pb}/${arch}/${branch}/tarballs/bindist.tar | awk '{print $4}')
-fi
+    . ${pb}/${arch}/portbuild.conf
+    . ${pb}/${arch}/portbuild.${node}
+
+    ${scp_cmd} -q -p ${pb}/scripts/setupnode ${client_user}@${node}:/tmp || return 1
+
+    client_setup="${ssh_cmd} -n ${client_user}@${node} sh /tmp/setupnode ${pb} ${arch} ${branch} ${buildid} ${scratchdir} \"${portsmd5}\" \"${srcmd5}\" \"${bindistmd5}\""
+    args="${nocopy} ${force}"
+
+    ${client_setup} pre-copy ${args} || (echo "pre-copy for ${node} failed"; return 1)
+
+    if [ "${norsync}" -eq 0 ]; then
+
+	rsync ${rsync_gzip} -e "${ssh_cmd}" -r -l -p --delete ${pb}/scripts \
+	    ${client_user}@${node}:${pb}/
+	checkerror $? || (echo "Copying scripts to ${node} failed"; return 1)
+
+	rsync ${rsync_gzip} -e "${ssh_cmd}" -r -l -p --delete ${pb}/${arch}/portbuild* \
+	    ${client_user}@${node}:${pb}/${arch}
+	checkerror $? || (echo "copying portbuild.conf files to ${node} failed"; return 1)
 
-scp -p ${pb}/scripts/setupnode ${client_user}@${node}:/tmp
-ssh -n ${client_user}@${node} ${sudo_cmd} sh /tmp/setupnode ${nocopy} $(hostname) ${pb} ${arch} ${branch} ${scratchdir} ${md5} ${disconnected}
+	if [ -f "${pb}/${arch}/clients/bindist-${node}.tar" ]; then
+	    rsync ${rsync_gzip} -e "${ssh_cmd}" -r -L -p --delete \
+		${pb}/${arch}/clients/bindist-${node}.tar \
+		${client_user}@${node}:${pb}/${arch}/clients/
+	    checkerror $? || (echo "Copying bindist-${node}.tar to ${node} failed"; return 1)
+	else
+	    echo "Host customization file not found: ${pb}/${arch}/clients/bindist-${node}.tar"
+	    return 1
+	fi
+
+	if [ "${buildid}" != "-" ]; then
+	    rsync ${rsync_gzip} -e "${ssh_cmd}" -r -L -p \
+		${builddir}/ports-${buildid}.tbz \
+		${builddir}/ports-${buildid}.tbz.md5 \
+		${builddir}/src-${buildid}.tbz \
+		${builddir}/src-${buildid}.tbz.md5 \
+		${builddir}/bindist.tbz \
+		${builddir}/bindist.tbz.md5 \
+		${client_user}@${node}:${builddir}/ 
+	    checkerror $? || (echo "Copying build tarballs to ${node} failed"; return 1)
+	fi
+    fi
+
+    ${client_setup} post-copy ${args} || (echo "post-copy for ${node} failed"; return 1)
+
+    if [ "${queue}" -eq 1 ]; then
+	lockf -k ${pb}/${arch}/queue/.lock \
+	    ${pb}/scripts/pollmachine ${arch}/${node} -queue
+    fi
+    echo "setting up of $node ended at $(date)"
+}
+
+pbab=${pb}/${arch}/${branch}
+
+norsync=0
+queue=0
+
+while [ $# -ge 1 ]; do
+    case $1 in
+	-norsync|-nocopy)
+	    norsync=1
+	    nocopy=-nocopy
+	    ;;
+	-queue)
+	    queue=1
+	    ;;
+	-force)
+	    force=-force
+	    ;;
+    esac
+    shift
+done
+
+if [ "${norsync}" -eq 0 ]; then
+    if [ "${branch}" != "-" -a "${buildid}" != "-" ]; then
+	buildid=$(resolve ${pb} ${arch} ${branch} ${buildid})
+
+	if [ -z "${buildid}" ]; then
+	    echo "Invalid build ID ${buildid}"
+	    exit 1
+	fi
+
+	builddir=${pbab}/builds/${buildid}
+
+	if [ ! -f ${builddir}/ports-${buildid}.tbz.md5 ]; then
+	    echo "ports-${buildid}.tbz.md5 not found"
+	    exit 1
+	else
+	    portsmd5=$(awk '{print $4}' ${builddir}/ports-${buildid}.tbz.md5)
+	fi
+	
+	if [ ! -f ${builddir}/src-${buildid}.tbz.md5 ]; then
+	    echo "src-${buildid}.tbz.md5 not found"
+	    exit 1
+	else
+	    srcmd5=$(awk '{print $4}' ${builddir}/src-${buildid}.tbz.md5)
+	fi
+	
+	if [ ! -f ${builddir}/bindist.tbz.md5 ]; then
+	    echo "bindist.tbz.md5 not found"
+	    exit 1
+	else
+	    bindistmd5=$(awk '{print $4}' ${builddir}/bindist.tbz.md5)
+	fi
+    fi
+fi
 
-if [ "${disconnected}" = 1 -a "${norsync}" = 0 ]; then
-  rsync ${rsync_gzip} -r -l -p --delete ${pb}/scripts ${client_user}@${node}:${pb}/
-  rsync ${rsync_gzip} -r -l -p --delete ${pb}/${arch}/portbuild* ${client_user}@${node}:${pb}/${arch}
-  rsync ${rsync_gzip} -r -L -p --delete ${pb}/${arch}/${branch}/ports ${client_user}@${node}:${pb}/${arch}/${branch}
-  rsync ${rsync_gzip} -r -L -p --delete ${pb}/${arch}/${branch}/src ${client_user}@${node}:${pb}/${arch}/${branch}
-  rsync ${rsync_gzip} -r -L -p --delete ${pb}/${arch}/${branch}/tarballs/ ${client_user}@${node}:${pb}/${arch}/${branch}/tarballs/
+if [ "${nodelist}" = "all" ]; then
+    nodelist=$(cat ${pb}/${arch}/mlist)
 fi
 
-lockf -k ${pb}/${arch}/queue/.lock ${pb}/scripts/pollmachine ${arch} ${node} -queue
-echo "setting up of $node ended at $(date)"
+for node in ${nodelist}; do
+    setup ${node} &
+done
+wait