From dcc89e3ad61799c4d21249ff4cc7d24a1fa394b7 Mon Sep 17 00:00:00 2001
From: gholt <z-launchpad@brim.net>
Date: Thu, 25 Oct 2012 18:27:16 +0000
Subject: [PATCH] dispersion report option to output missing parts

Adds a -p option to swift-dispersion-report that will output the
partitions missing copies to standard error. Another thing we've been
meaning to add for forever. It's useful when you want to do some
further research on whether a partition truly has fewer copies or if
they're just somewhere else in the cluster due to a ring change, for
instance.

DocImpact

Change-Id: I5e47aa5818483ecc34b39ef6f8cd83ad312ed9a0
---
 bin/swift-dispersion-report            | 30 ++++++++++++++++++++------
 doc/manpages/swift-dispersion-report.1 | 14 +++++++++++-
 2 files changed, 36 insertions(+), 8 deletions(-)

diff --git a/bin/swift-dispersion-report b/bin/swift-dispersion-report
index 4ff4468226..b98f6dd664 100755
--- a/bin/swift-dispersion-report
+++ b/bin/swift-dispersion-report
@@ -66,7 +66,7 @@ def get_error_log(prefix):
 
 
 def container_dispersion_report(coropool, connpool, account, container_ring,
-                                retries):
+                                retries, output_missing_partitions):
     with connpool.item() as conn:
         containers = [c['name'] for c in conn.get_account(prefix='dispersion_',
                                                          full_listing=True)[1]]
@@ -100,6 +100,13 @@ def container_dispersion_report(coropool, connpool, account, container_ring,
             except (Exception, Timeout), err:
                 error_log('Giving up on /%s/%s/%s: %s' % (part, account,
                           container, err))
+        if output_missing_partitions and \
+                found_count < container_ring.replica_count:
+            missing = container_ring.replica_count - found_count
+            print '\r\x1B[K',
+            stdout.flush()
+            print >>stderr, '# Container partition %s missing %s cop%s' % (
+                part, missing, 'y' if missing == 1 else 'ies')
         container_copies_found[found_count] += 1
         containers_queried[0] += 1
         if time() >= next_report[0]:
@@ -155,7 +162,7 @@ def container_dispersion_report(coropool, connpool, account, container_ring,
 
 
 def object_dispersion_report(coropool, connpool, account, object_ring,
-                             retries):
+                             retries, output_missing_partitions):
     container = 'dispersion_objects'
     with connpool.item() as conn:
         try:
@@ -198,6 +205,13 @@ def object_dispersion_report(coropool, connpool, account, object_ring,
             except (Exception, Timeout), err:
                 error_log('Giving up on /%s/%s/%s/%s: %s' % (part, account,
                           container, obj, err))
+        if output_missing_partitions and \
+                found_count < object_ring.replica_count:
+            missing = object_ring.replica_count - found_count
+            print '\r\x1B[K',
+            stdout.flush()
+            print >>stderr, '# Object partition %s missing %s cop%s' % (
+                part, missing, 'y' if missing == 1 else 'ies')
         object_copies_found[found_count] += 1
         objects_queried[0] += 1
         if time() >= next_report[0]:
@@ -289,6 +303,8 @@ Usage: %prog [options] [conf_file]
                       help='dump dispersion report in json format')
     parser.add_option('-d', '--debug', action='store_true', default=False,
                       help='print 404s to standard error')
+    parser.add_option('-p', '--partitions', action='store_true', default=False,
+                      help='print missing partitions to standard error')
 
     options, args = parser.parse_args()
 
@@ -324,11 +340,11 @@ Usage: %prog [options] [conf_file]
     container_ring = Ring(swift_dir, ring_name='container')
     object_ring = Ring(swift_dir, ring_name='object')
 
-    container_result = container_dispersion_report(coropool, connpool,
-                                                   account, container_ring,
-                                                   retries)
-    object_result = object_dispersion_report(coropool, connpool, account,
-                                             object_ring, retries)
+    container_result = container_dispersion_report(
+        coropool, connpool, account, container_ring, retries,
+        options.partitions)
+    object_result = object_dispersion_report(
+        coropool, connpool, account, object_ring, retries, options.partitions)
     if json_output:
         print json.dumps({"container": container_result,
                           "object": object_result})
diff --git a/doc/manpages/swift-dispersion-report.1 b/doc/manpages/swift-dispersion-report.1
index 785fc5f7fb..414e280d63 100644
--- a/doc/manpages/swift-dispersion-report.1
+++ b/doc/manpages/swift-dispersion-report.1
@@ -24,7 +24,7 @@
 
 .SH SYNOPSIS
 .LP
-.B swift-dispersion-report [-j|--dump-json] [conf_file]
+.B swift-dispersion-report [-d|--debug] [-j|--dump-json] [-p|--partitions] [conf_file]
 
 .SH DESCRIPTION 
 .PP
@@ -54,12 +54,24 @@ same configuration file, /etc/swift/dispersion.conf . The account used by these
 tool should be a dedicated account for the dispersion stats and also have admin
 privileges. 
 
+.SH OPTIONS
+.RS 0
+.PD 1
+.IP "\fB-d, --debug\fR"
+output any 404 responses to standard error
+
 .SH OPTIONS
 .RS 0
 .PD 1
 .IP "\fB-j, --dump-json\fR"
 output dispersion report in json format
 
+.SH OPTIONS
+.RS 0
+.PD 1
+.IP "\fB-p, --partitions\fR"
+output the partition numbers that have any missing replicas
+
 .SH CONFIGURATION
 .PD 0 
 Example \fI/etc/swift/dispersion.conf\fR: