[trafficserver] 03/08: TS-4870: Avoid marking storage offline multiple times

[bc...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

bcall pushed a commit to branch 7.0.x
in repository https://git-dual.apache.org/repos/asf/trafficserver.git

commit 7ab269cc245b0da6c78da5e187ea2ad2b1b8365b
Author: Gancho Tenev <gt...@gmail.com>
AuthorDate: Thu Sep 15 06:44:44 2016 -0700

    TS-4870: Avoid marking storage offline multiple times
    
    Added a CacheDisk:online flag because we cannot relay on DISK_BAD(disk)
    macro to identify if we already marked it bad. The problem is that in the
    common use-case in handle_disk_failure(), the disk is already bad when
    mark_storage_offline() is called, so we can't depend on the good->bad state
    transition to know when to update the accounting.
    
    (cherry picked from commit 7421eabf0f74da0005c25cc199b79f28ff5159de)
---
 iocore/cache/Cache.cc      | 9 ++++++++-
 iocore/cache/P_CacheDisk.h | 2 ++
 2 files changed, 10 insertions(+), 1 deletion(-)

diff --git a/iocore/cache/Cache.cc b/iocore/cache/Cache.cc
index e4c5676..9562808 100644
--- a/iocore/cache/Cache.cc
+++ b/iocore/cache/Cache.cc
@@ -2000,6 +2000,13 @@ CacheProcessor::mark_storage_offline(CacheDisk *d ///< Target disk
   uint64_t total_dir_delete   = 0;
   uint64_t used_dir_delete    = 0;
 
+  /* Don't mark it again, it will invalidate the stats! */
+  if (!d->online) {
+    return this->has_online_storage();
+  }
+
+  d->online = false;
+
   if (!DISK_BAD(d))
     SET_DISK_BAD(d);
 
@@ -2052,7 +2059,7 @@ CacheProcessor::has_online_storage() const
 {
   CacheDisk **dptr = gdisks;
   for (int disk_no = 0; disk_no < gndisks; ++disk_no, ++dptr) {
-    if (!DISK_BAD(*dptr))
+    if (!DISK_BAD(*dptr) && (*dptr)->online)
       return true;
   }
   return false;
diff --git a/iocore/cache/P_CacheDisk.h b/iocore/cache/P_CacheDisk.h
index b391625..8d631e8 100644
--- a/iocore/cache/P_CacheDisk.h
+++ b/iocore/cache/P_CacheDisk.h
@@ -97,6 +97,7 @@ struct CacheDisk : public Continuation {
   int num_errors;
   int cleared;
   bool read_only_p;
+  bool online; /* flag marking cache disk online or offline (because of too many failures or by the operator). */
 
   // Extra configuration values
   int forced_volume_num;           ///< Volume number for this disk.
@@ -119,6 +120,7 @@ struct CacheDisk : public Continuation {
       num_errors(0),
       cleared(0),
       read_only_p(false),
+      online(true),
       forced_volume_num(-1)
   {
   }

-- 
To stop receiving notification emails like this one, please contact
"commits@trafficserver.apache.org" <co...@trafficserver.apache.org>.