Only restore settings when -c given explicitly

The previous behavior would detect if the appliance had ever had a
manual config run and automatically restore settings if not. This
has caused a number of problems in manual testing, some due to bugs
importing settings from an 11.10.34x appliance into a v2.0 appliance
and others due to settings like hostname and SSL being imported into
appliances that are being set up as secondary / staging servers.
It's also just been surprising that settings aren't restored on
subsequent runs.

Making the settings restore an option that must be given explicitly
removes much of the confusion around the auto detection at the
expense of a slightly smoother experience in the perfect case.

Author: rtomayko

bin/ghe-restore

@@ -6,10 +6,10 @@
 #/ argument is provided, it always overrides the configured restore host.
 #/
 #/ Options:
-#/   -c                Clean restore. Restores appliance settings and license
-#/                     in addition to datastores. This is the default for
-#/                     appliances that have not yet been configured but can be
-#/                     forced by passing the argument explicitly.
+#/   -c                Restore appliance settings and license in addition to
+#/                     datastores. Settings are not restored by default to
+#/                     prevent overwriting different configuration on the
+#/                     restore host.
 #/   -s <snapshot-id>  Restore from the snapshot with the given id. Available
 #/                     snapshots may be listed under the data directory.
 #/   -v                Enable verbose output.
@@ -31,15 +31,15 @@ GHE_RESTORE_SNAPSHOT="current"
 export GHE_RESTORE_SNAPSHOT
 
 # Parse arguments
-clean_restore=false
+restore_settings=false
 while true; do
     case "$1" in
         -s)
             GHE_RESTORE_SNAPSHOT="$(basename "$2")"
             shift 2
             ;;
         -c)
-            clean_restore=true
+            restore_settings=true
             shift
             ;;
         -*)
@@ -55,6 +55,9 @@ done
 # Grab the host arg
 host="${1:-$GHE_RESTORE_HOST}"
 
+# Hostname without any port suffix
+hostname=$(echo "$host" | cut -f 1 -d :)
+
 # Show usage with no <host>
 [ -z "$host" ] && print_usage
 
@@ -72,25 +75,29 @@ echo "Starting $GHE_BACKUP_STRATEGY restore of $host from snapshot $GHE_RESTORE_
 # Perform a host-check and establish the remote version in GHE_REMOTE_VERSION.
 ghe_remote_version_required "$host"
 
-# Figure out if host has ever been configured based on the presence of the
-# dna.json file. If not, enable restoring settings and license in addition
-# to primary datastores.
-if ! $clean_restore &&
+# Verify the host has been fully configured at least once unless the -c
+# argument was provided.
+if ! $restore_settings &&
    ! ghe-ssh "$host" -- "test -f '$GHE_REMOTE_DATA_DIR/enterprise/dna.json' -o \
                               -f '$GHE_REMOTE_DATA_USER_DIR/common/dna.json'"; then
-    clean_restore=true
+    echo "Error: $hostname not configured." 1>&2
+    echo "Please visit https://$hostname/setup/settings to configure base appliance settings before continuing." 1>&2
+    exit 1
 fi
 
 # Make sure the GitHub appliance is in maintenance mode and all writing
 # processes have bled out.
 ghe-maintenance-mode-enable "$host"
 
-# Restore settings and license if performing a clean restore.
-if $clean_restore; then
+# Restore settings and license if requested.
+if $restore_settings; then
     ghe-restore-settings "$host"
+fi
 
-    # Make sure mysql and elasticsearch are setup and running since this is a
-    # fresh install which hasn't had a configure run yet.
+# Make sure mysql and elasticsearch are prep'd and running before restoring into
+# appliances v2.x or greater. These services will not have been started on appliances
+# that have not been configured yet.
+if [ "$GHE_VERSION_MAJOR" -ge 2 ]; then
     echo "sudo ghe-service-ensure-mysql && sudo ghe-service-ensure-elasticsearch" |
     ghe-ssh "$host" -- /bin/sh 1>&3
 fi
@@ -132,4 +139,4 @@ echo "Restoring SSH host keys ..."
 ghe-ssh "$host" -- 'ghe-import-ssh-host-keys' < "$GHE_RESTORE_SNAPSHOT_PATH/ssh-host-keys.tar" 1>&3
 
 echo "Completed restore of $host from snapshot $GHE_RESTORE_SNAPSHOT"
-echo "Visit https://$(echo "$host" | cut -f 1 -d :)/setup/settings in a browser to configure the recovered appliance."
+echo "Visit https://$hostname/setup/settings to configure the recovered appliance."

test/test-ghe-restore.sh

@@ -63,21 +63,22 @@ echo "fake license data" > "$GHE_DATA_DIR/current/enterprise.ghl"
 echo "fake manage password hash data" > "$GHE_DATA_DIR/current/manage-password"
 echo "rsync" > "$GHE_DATA_DIR/current/strategy"
 
-begin_test "ghe-restore into unconfigured vm"
+begin_test "ghe-restore into configured vm"
 (
     set -e
     rm -rf "$GHE_REMOTE_DATA_DIR"
     setup_remote_metadata
 
+    # create settings file -- used to determine if instance has been configured.
+    touch "$GHE_REMOTE_DATA_DIR/enterprise/dna.json"
+
     # set restore host environ var
     GHE_RESTORE_HOST=127.0.0.1
     export GHE_RESTORE_HOST
 
     # run ghe-restore and write output to file for asserting against
-    if ! ghe-restore -v > "$TRASHDIR/restore-out" 2>&1; then
-        cat "$TRASHDIR/restore-out"
-        false
-    fi
+    ghe-restore -v > "$TRASHDIR/restore-out" 2>&1
+    cat "$TRASHDIR/restore-out"
 
     # verify connect to right host
     grep -q "Connect 127.0.0.1 OK" "$TRASHDIR/restore-out"
@@ -88,27 +89,10 @@ begin_test "ghe-restore into unconfigured vm"
     grep -q "fake ghe-export-redis data" "$TRASHDIR/restore-out"
     grep -q "fake ghe-export-authorized-keys data" "$TRASHDIR/restore-out"
     grep -q "fake ghe-export-ssh-host-keys data" "$TRASHDIR/restore-out"
-    grep -q "fake ghe-export-settings data" "$TRASHDIR/restore-out"
 
-    # verify ghe-import-es-indices is run under 1.x VMs and that the
-    # elasticsearch-legacy directory was created under 2.x VMs.
-    if [ "$GHE_VERSION_MAJOR" -eq 1 ]; then
-        grep -q "ghe-import-es-indices" "$TRASHDIR/restore-out"
-    elif [ "$GHE_VERSION_MAJOR" -ge 2 ]; then
-        test -d "$GHE_REMOTE_DATA_USER_DIR/elasticsearch-legacy"
-    fi
-
-    # verify manage password was restored under v2.x or greater VMs
-    if [ "$GHE_VERSION_MAJOR" -ge 2 ]; then
-        test -f "$GHE_REMOTE_DATA_USER_DIR/common/manage-password"
-        [ "$(cat "$GHE_REMOTE_DATA_USER_DIR/common/manage-password")" = "fake manage password hash data" ]
-    fi
-
-    # verify service-ensure scripts were run under versions >= v2.x
-    if [ "$GHE_VERSION_MAJOR" -ge 2 ]; then
-        grep -q "ghe-service-ensure-mysql OK" "$TRASHDIR/restore-out"
-        grep -q "ghe-service-ensure-elasticsearch OK" "$TRASHDIR/restore-out"
-    fi
+    # verify settings import was *not* run due to instance already being
+    # configured.
+    ! grep -q "fake ghe-export-settings data" "$TRASHDIR/restore-out"
 
     # verify all repository data was transferred to the restore location
     diff -ru "$GHE_DATA_DIR/current/repositories" "$GHE_REMOTE_DATA_USER_DIR/repositories"
@@ -126,21 +110,18 @@ begin_test "ghe-restore into unconfigured vm"
 )
 end_test
 
-begin_test "ghe-restore into configured vm"
+begin_test "ghe-restore -c into unconfigured vm"
 (
     set -e
     rm -rf "$GHE_REMOTE_DATA_DIR"
     setup_remote_metadata
 
-    # create settings file -- used to determine if instance has been configured.
-    touch "$GHE_REMOTE_DATA_DIR/enterprise/dna.json"
-
     # set restore host environ var
     GHE_RESTORE_HOST=127.0.0.1
     export GHE_RESTORE_HOST
 
     # run ghe-restore and write output to file for asserting against
-    ghe-restore -v > "$TRASHDIR/restore-out" 2>&1
+    ghe-restore -v -c > "$TRASHDIR/restore-out" 2>&1
     cat "$TRASHDIR/restore-out"
 
     # verify connect to right host
@@ -153,9 +134,8 @@ begin_test "ghe-restore into configured vm"
     grep -q "fake ghe-export-authorized-keys data" "$TRASHDIR/restore-out"
     grep -q "fake ghe-export-ssh-host-keys data" "$TRASHDIR/restore-out"
 
-    # verify settings import was *not* run due to instance already being
-    # configured.
-    ! grep -q "fake ghe-export-settings data" "$TRASHDIR/restore-out"
+    # verify settings were imported
+    grep -q "fake ghe-export-settings data" "$TRASHDIR/restore-out"
 
     # verify all repository data was transferred to the restore location
     diff -ru "$GHE_DATA_DIR/current/repositories" "$GHE_REMOTE_DATA_USER_DIR/repositories"
@@ -173,49 +153,22 @@ begin_test "ghe-restore into configured vm"
 )
 end_test
 
-begin_test "ghe-restore -c into configured vm"
+begin_test "ghe-restore into unconfigured vm"
 (
     set -e
     rm -rf "$GHE_REMOTE_DATA_DIR"
     setup_remote_metadata
 
-    # create settings file -- used to determine if instance has been configured.
-    touch "$GHE_REMOTE_DATA_DIR/enterprise/dna.json"
-
     # set restore host environ var
     GHE_RESTORE_HOST=127.0.0.1
     export GHE_RESTORE_HOST
 
     # run ghe-restore and write output to file for asserting against
-    ghe-restore -v -c > "$TRASHDIR/restore-out" 2>&1
-    cat "$TRASHDIR/restore-out"
-
-    # verify connect to right host
-    grep -q "Connect 127.0.0.1 OK" "$TRASHDIR/restore-out"
-
-    # verify all import scripts were run
-    grep -q "alice/index.html" "$TRASHDIR/restore-out"
-    grep -q "fake ghe-export-mysql data" "$TRASHDIR/restore-out"
-    grep -q "fake ghe-export-redis data" "$TRASHDIR/restore-out"
-    grep -q "fake ghe-export-authorized-keys data" "$TRASHDIR/restore-out"
-    grep -q "fake ghe-export-ssh-host-keys data" "$TRASHDIR/restore-out"
+    # this should fail due to the appliance being in an unconfigured state
+    ! ghe-restore -v > "$TRASHDIR/restore-out" 2>&1
 
-    # verify settings were imported
-    grep -q "fake ghe-export-settings data" "$TRASHDIR/restore-out"
-
-    # verify all repository data was transferred to the restore location
-    diff -ru "$GHE_DATA_DIR/current/repositories" "$GHE_REMOTE_DATA_USER_DIR/repositories"
-
-    # verify all pages data was transferred to the restore location
-    diff -ru "$GHE_DATA_DIR/current/pages" "$GHE_REMOTE_DATA_USER_DIR/pages"
-
-    if [ "$GHE_VERSION_MAJOR" -ge 2 ]; then
-        # verify all hookshot user data was transferred
-        diff -ru "$GHE_DATA_DIR/current/hookshot" "$GHE_REMOTE_DATA_USER_DIR/hookshot"
-
-        # verify all alambic assets user data was transferred
-        diff -ru "$GHE_DATA_DIR/current/alambic_assets" "$GHE_REMOTE_DATA_USER_DIR/alambic_assets"
-    fi
+    # verify that ghe-restore failed due to the appliance not being configured
+    grep -q -e "Error: $GHE_RESTORE_HOST not configured" "$TRASHDIR/restore-out"
 )
 end_test
 
@@ -225,6 +178,9 @@ begin_test "ghe-restore with host arg"
     rm -rf "$GHE_REMOTE_DATA_DIR"
     setup_remote_metadata
 
+    # create settings file -- used to determine if instance has been configured.
+    touch "$GHE_REMOTE_DATA_DIR/enterprise/dna.json"
+
     # set restore host environ var
     GHE_RESTORE_HOST=127.0.0.1
     export GHE_RESTORE_HOST
@@ -257,6 +213,9 @@ begin_test "ghe-restore no host arg or configured restore host"
     rm -rf "$GHE_REMOTE_DATA_DIR"
     setup_remote_metadata
 
+    # create settings file -- used to determine if instance has been configured.
+    touch "$GHE_REMOTE_DATA_DIR/enterprise/dna.json"
+
     # unset configured restore host
     unset GHE_RESTORE_HOST
 
@@ -271,6 +230,9 @@ begin_test "ghe-restore with no pages backup"
     rm -rf "$GHE_REMOTE_DATA_DIR"
     setup_remote_metadata
 
+    # create settings file -- used to determine if instance has been configured.
+    touch "$GHE_REMOTE_DATA_DIR/enterprise/dna.json"
+
     # remove pages data
     rm -rf "$GHE_DATA_DIR/1/pages"
 
@@ -285,6 +247,9 @@ begin_test "ghe-restore with tarball strategy"
     rm -rf "$GHE_REMOTE_DATA_DIR"
     setup_remote_metadata
 
+    # create settings file -- used to determine if instance has been configured.
+    touch "$GHE_REMOTE_DATA_DIR/enterprise/dna.json"
+
     # run it
     echo "tarball" > "$GHE_DATA_DIR/current/strategy"
     output=$(ghe-restore -v localhost)