[
https://issues.apache.org/jira/browse/TS-4331?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15304211#comment-15304211
]
ASF GitHub Bot commented on TS-4331:
------------------------------------
Github user jpeach commented on a diff in the pull request:
https://github.com/apache/trafficserver/pull/653#discussion_r64924636
--- Diff: iocore/hostdb/P_RefCountCache.h ---
@@ -0,0 +1,572 @@
+#ifndef _P_RefCountCache_h_
+#define _P_RefCountCache_h_
+
+#include "I_EventSystem.h"
+
+#include <ts/Map.h>
+#include <ts/List.h>
+
+#include <cstdint>
+// TODO: no vector?
+#include <vector>
+#include <unistd.h>
+#include <string.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <sys/mman.h>
+#include <cmath>
+#include <fstream>
+
+#include "ts/I_Version.h"
+
+#define REFCOUNT_CACHE_EVENT_SYNC REFCOUNT_CACHE_EVENT_EVENTS_START
+
+#define REFCOUNTCACHE_MAGIC_NUMBER 0x0BAD2D9
+#define REFCOUNTCACHE_MAJOR_VERSION 1
+#define REFCOUNTCACHE_MINOR_VERSION 0
+
+// TODO: elsewhere?
+#define PtrMutex Ptr<ProxyMutex>
+
+// In the cache we want to recount our items
+// This class will handle memory allocations (on the freelist) and
+// take care of implementing RefCountObj-- so that the items we want in
+// the cache don't have to inherit from it (which makes
serialization/deserialization much easier)
+class RefCountCacheItemBase : public RefCountObj
+{
+public:
+ RefCountCacheItemBase(uint64_t key = 0, unsigned int size = 0, char
*iobuf = NULL, int iobuffer_index = 0)
+ {
+ this->key = key;
+ this->size = size;
+ this->iobuf = iobuf;
+ this->iobuffer_index = iobuffer_index;
+ }
+ void
+ free()
+ {
+ if (this->size > 0) {
+ // free the block
+ ioBufAllocator[iobuffer_index].free_void((void *)(iobuf));
+ }
+ }
+ uint64_t key; // Key of this item
+ unsigned int size; // how much space does this guy get
+ int64_t iobuffer_index;
+ char *iobuf;
+};
+
+// Template to the particular class, so we can caste the item ptr correctly
+template <class C> class RefCountCacheItem : public RefCountCacheItemBase
+{
+public:
+ using RefCountCacheItemBase::RefCountCacheItemBase;
+ C *item();
+};
+
+template <class C>
+C *
+RefCountCacheItem<C>::item()
+{
+ return (C *)this->iobuf;
+}
+
+// Layer of inderection for the hashmap-- since it needs lots of things
inside of it
+class RefCountCacheHashingValue
+{
+public:
+ Ptr<RefCountCacheItemBase> item;
+ LINK(RefCountCacheHashingValue, item_link);
+ RefCountCacheHashingValue(Ptr<RefCountCacheItemBase> i) : item(i){};
+};
+
+// TODO: template?
+struct RefCountCacheHashing {
+ typedef uint64_t ID;
+ typedef uint64_t const Key;
+ typedef RefCountCacheHashingValue Value;
+ typedef DList(RefCountCacheHashingValue, item_link) ListHead;
+
+ static ID
+ hash(Key key)
+ {
+ return key;
+ }
+ static Key
+ key(Value const *value)
+ {
+ return value->item->key;
+ }
+ static bool
+ equal(Key lhs, Key rhs)
+ {
+ return lhs == rhs;
+ }
+};
+
+// The RefCountCachePartition is simply a map of key ->
Ptr<RefCountCacheItem<YourClass>>
+// TODO: add metrics
+// TODO: store partition number for debugging
+template <class C> class RefCountCachePartition
+{
+public:
+ RefCountCachePartition(int maxSize, int maxItems);
+ RefCountCacheItem<C> *alloc(uint64_t key, int size = 0);
+ Ptr<RefCountCacheItem<C>> get(uint64_t key);
+ void del(uint64_t key);
+ void clear();
+
+ int numItems();
+ void copy(std::vector<Ptr<RefCountCacheItemBase>> *items);
+
+ typedef typename TSHashTable<RefCountCacheHashing>::iterator
iteratortype;
+ typedef typename TSHashTable<RefCountCacheHashing>::self hashtype;
+ typedef typename TSHashTable<RefCountCacheHashing>::Location
locationtype;
+ TSHashTable<RefCountCacheHashing> *getMap();
+
+ PtrMutex lock; // Lock
+
+private:
+ int maxSize;
+ int maxItems;
+
+ int size;
+ int items;
+
+ hashtype itemMap;
+};
+
+template <class C> RefCountCachePartition<C>::RefCountCachePartition(int
maxSize, int maxItems)
+{
+ this->maxSize = maxSize;
+ this->maxItems = maxItems;
+ this->size = 0;
+ this->items = 0;
+
+ // Initialize lock
+ this->lock = new_ProxyMutex();
+}
+
+// TODO: error somehow if you asked for something too big!
+// Allocate space for an item `C` as well as `size` extra space.
+template <class C>
+RefCountCacheItem<C> *
+RefCountCachePartition<C>::alloc(uint64_t key, int size)
+{
+ size += sizeof(C);
+ // Remove any colliding entries
+ this->del(key);
+
+ // Now allocat a sufficiently large iobuf to store our data
+ // as well as the intermediate layers we have
+ int allocSize = size + sizeof(RefCountCacheItem<C>) +
sizeof(RefCountCacheHashingValue);
+ // allocate a sufficiently large block from the ioBufAllocator
+ int iobuffer_index = iobuffer_size_to_index(allocSize);
+ ink_release_assert(iobuffer_index >= 0);
+ char *iobuf = (char *)ioBufAllocator[iobuffer_index].alloc_void();
+
+ RefCountCacheItem<C> *item = new (iobuf + size)
RefCountCacheItem<C>(key, size, iobuf, iobuffer_index);
+
+ // TODO: either put the RefCountCacheItem on the same iobuf, or move this
+ // to a class allocator
+
+ // if we are past our limits, then lets not store a reference to the
return in our map
+ if ((this->maxItems > 0 && this->items > this->maxItems) ||
(this->maxSize > 0 && this->size > this->maxSize)) {
+ return item;
+ }
+
+ // Add our entry to the map
+ RefCountCacheHashingValue *val =
+ new ((char *)item + sizeof(RefCountCacheItem<C>))
RefCountCacheHashingValue(make_ptr((RefCountCacheItemBase *)item));
+ this->itemMap.insert(val);
+
+ return item;
+}
+
+template <class C>
+Ptr<RefCountCacheItem<C>>
+RefCountCachePartition<C>::get(uint64_t key)
+{
+ locationtype l = this->itemMap.find(key);
+ if (l.isValid()) {
+ // found
+ return make_ptr((RefCountCacheItem<C> *)l.m_value->item.get());
+ } else {
+ return Ptr<RefCountCacheItem<C>>(NULL);
+ }
+}
+
+template <class C>
+void
+RefCountCachePartition<C>::del(uint64_t key)
+{
+ this->itemMap.remove(key);
+}
+
+template <class C>
+void
+RefCountCachePartition<C>::clear()
+{
+ // Clear the in memory hashmap
+ // TODO: delete all items (not sure if clear calls delete on all items)
+ this->itemMap.clear();
+ this->items = 0;
+ this->size = 0;
+}
+
+template <class C>
+int
+RefCountCachePartition<C>::numItems()
+{
+ return this->itemMap.count();
+}
+
+template <class C>
+void
+RefCountCachePartition<C>::copy(std::vector<Ptr<RefCountCacheItemBase>>
*items)
+{
+ for (RefCountCachePartition<C>::iteratortype i = this->itemMap.begin();
i != this->itemMap.end(); ++i) {
+ items->push_back(make_ptr((RefCountCacheItemBase *)i->item.get()));
+ }
+}
+
+// TODO: pass an iterator or something? Seems not good to pass a pointer
to a member
+template <class C>
+TSHashTable<RefCountCacheHashing> *
+RefCountCachePartition<C>::getMap()
+{
+ return &this->itemMap;
+}
+
+// The implementation of this class must be in here, as this is used by
template
+// classes, and c++ rocks that way
+// The header for the cache, this is used to check if the serialized cache
is compatible
+class RefCountCacheHeader
+{
+public:
+ unsigned int magic;
+ VersionNumber version;
+
+ int numPartitions;
+ int partitionMaxSize;
+
+ RefCountCacheHeader(int numParts, int partSize)
+ : magic(REFCOUNTCACHE_MAGIC_NUMBER), numPartitions(numParts),
partitionMaxSize(partSize)
+ {
+ this->version.ink_major = REFCOUNTCACHE_MAJOR_VERSION;
+ this->version.ink_minor = REFCOUNTCACHE_MINOR_VERSION;
+ };
+
+ bool
+ operator==(const RefCountCacheHeader other) const
+ {
+ return (this->magic == other.magic && this->version.ink_major ==
other.version.ink_major &&
+ this->version.ink_minor == other.version.ink_minor &&
this->numPartitions == other.numPartitions &&
+ this->partitionMaxSize == other.partitionMaxSize);
+ }
+
+ bool
+ compatible(RefCountCacheHeader *other)
+ {
+ // printf("magic %d %d\n", this->magic, other->magic);
+ // printf("version.ink_major %d %d\n", this->version.ink_major,
other->version.ink_major);
+ // printf("numPartitions %d %d\n", this->numPartitions,
other->numPartitions);
+ // printf("partitionMaxSize %d %d\n", this->partitionMaxSize,
other->partitionMaxSize);
+
+ return (this->magic == other->magic && this->version.ink_major ==
other->version.ink_major &&
+ this->numPartitions == other->numPartitions &&
this->partitionMaxSize == other->partitionMaxSize);
+ };
+};
+
+// Base for syncing purposes
+// Since the actual RefCountCache is templated, we don't want to require
+// importing the template type for all uses (such as syncing)
+class RefCountCacheBase
--- End diff --
Using ``Base`` as a base class name is fairly ugly. This isn't really
related to the cache interface at all, it is a persistence interface.
All the virtual methods ought to be pure virtual. All methods that don't
mutate the object should be ``const``.
> Hostdb consistency problems due to MultiCache
> ---------------------------------------------
>
> Key: TS-4331
> URL: https://issues.apache.org/jira/browse/TS-4331
> Project: Traffic Server
> Issue Type: Bug
> Components: HostDB
> Reporter: Thomas Jackson
> Assignee: Thomas Jackson
> Fix For: 7.0.0
>
>
> This ticket is for the correct long term fix to TS-4207
> pulled from a comment, which wraps up the issue
> {quote}
> Leif Hedstrom I have spent a decent amount of time on this while I was OOO on
> vacation the last couple of weeks. It seems that the root cause of this issue
> has always existed, and that the addition of always doing hostname storing
> (https://github.com/apache/trafficserver/commit/0e703e1e) we are just causing
> the issue to happen all the time.
> To understand the issue I'll give a little background in how hostdb is
> currently working. Basically hostdb is just a wrapper around this templated
> struct called MultiCache. MultiCache is "multi" not because it is templated,
> but because it has two types of storage (static-- blocks and dynamic--
> alloc). The static side of the cache can hold N HostDBInfo structs (the
> results of DNS queries). The dynamic side is used to store the round robin
> records and various strings associated with the record. The size of this
> dynamic space is defined as (N x [estimated_heap_bytes_per_entry. The basic
> problem we are running into is that we are putting too much preassure on the
> dynamic heap-- such that the heap is getting re-used while people still have
> references to items in that space.
> So, I've actually been working on re-writing MultiCache to allocate the
> entire required block at once (so we don't have this problem where the parent
> exists but not the children), but I'm not certain if we want such a change to
> go into the 6.x branch (I'm willing to discuss if we want). If we aren't
> comfortable with such a large change I suggest just accounting for the
> hostname size in the estimated_heap_bytes_per_entry as a stopgap solution.
> The maximum allowable size is 253 (so 254 with null terminator), but we could
> pick a smaller number (~120 or so seems to be more reasonable). Alternatively
> you can increase the number of records in hostdb (and the size accordingly)
> to increase the dynamic heap size.
> TLDR; almost done with the long term solution, but I'm not sure if we want to
> merge that into 6.x-- alternatively we can do a simple workaround in 6.x
> (https://github.com/apache/trafficserver/pull/553)
> {quote}
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
