Introducted by #13013
After defragmenting the dictionary in the kvstore, if the dict is
reallocated, the value of its node in the kvstore rehashing list must be
updated.
## Background
1. Currently Lua memory control does not pass through Redis's zmalloc.c.
Redis maxmemory cannot limit memory problems caused by users abusing lua
since these lua VM memory is not part of used_memory.
2. Since jemalloc is much better (fragmentation and speed), and also we
know it and trust it. we are
going to use jemalloc instead of libc to allocate the Lua VM code and
count it used memory.
## Process:
In this PR, we will use jemalloc in lua.
1. Create an arena for all lua vm (script and function), which is
shared, in order to avoid blocking defragger.
2. Create a bound tcache for the lua VM, since the lua VM and the main
thread are by default in the same tcache, and if there is no isolated
tcache, lua may request memory from the tcache which has just been freed
by main thread, and vice versa
On the other hand, since lua vm might be release in bio thread, but
tcache is not thread-safe, we need to recreate
the tcache every time we recreate the lua vm.
3. Remove lua memory statistics from memory fragmentation statistics to
avoid the effects of lua memory fragmentation
## Other
Add the following new fields to `INFO DEBUG` (we may promote them to
INFO MEMORY some day)
1. allocator_allocated_lua: total number of bytes allocated of lua arena
2. allocator_active_lua: total number of bytes in active pages allocated
in lua arena
3. allocator_resident_lua: maximum number of bytes in physically
resident data pages mapped in lua arena
4. allocator_frag_bytes_lua: fragment bytes in lua arena
This is oranagra's idea, and i got some help from sundb.
This solves the third point in #13102.
---------
Co-authored-by: debing.sun <debing.sun@redis.com>
Co-authored-by: Oran Agra <oran@redislabs.com>
After #13013
### This PR make effort to defrag the pubsub kvstore in the following
ways:
1. Till now server.pubsub(shard)_channels only share channel name obj
with the first subscribed client, now change it so that the clients and
the pubsub kvstore share the channel name robj.
This would save a lot of memory when there are many subscribers to the
same channel.
It also means that we only need to defrag the channel name robj in the
pubsub kvstore, and then update
all client references for the current channel, avoiding the need to
iterate through all the clients to do the same things.
2. Refactor the code to defragment pubsub(shard) in the same way as
defragment of keys and EXPIRES, with the exception that we only
defragment pubsub(without shard) when slot is zero.
### Other
Fix an overlook in #11695, if defragment doesn't reach the end time, we
should wait for the current
db's keys and expires, pubsub and pubsubshard to finish before leaving,
now it's possible to exit
early when the keys are defragmented.
---------
Co-authored-by: oranagra <oran@redislabs.com>
Implement #12963
## Changes
1. large bins don't have external fragmentation or are at least
non-defraggable, so we should ignore the effect of
large bins when measuring fragmentation, and only measure fragmentation
of small bins. this affects both the allocator_frag* metrics and also
the active-defrag trigger
2. Adding INFO metrics for `muzzy` memory, which is memory returned to
the OS but still shows as RSS until the OS reclaims it.
---------
Co-authored-by: Oran Agra <oran@redislabs.com>
Fail CI:
https://github.com/redis/redis/actions/runs/7837608438/job/21387609715
## Why defragment tests only failed under 32-bit
First of all, under 32-bit jemalloc will allocate more small bins and
less large bins, which will also lead to more external fragmentation,
therefore, the fragmentation ratio is higher in 32-bit than in 64-bit,
so the defragment tests(`Active defrag eval scripts: cluster` and
`Active defrag big keys: cluster`) always fails in 32-bit.
## Why defragment tests only failed with cluster
The fowllowing is the result of `Active defrag eval scripts: cluster`
test.
1) Before #11695, the fragmentation ratio is 3.11%.
2) After #11695, the fragmentation ratio grew to 4.58%.
Since we are using per-slot dictionary to manage slots, we will only
defragment the contents of these dictionaries (keys, values), but not
the dictionaries' struct and ht_table, which means that frequent
shrinking and expanding of the dictionaries, will make more fragments.
3) After #12850 and #12948, In cluster mode, a large number of cluster
slot dicts will be shrunk, creating additional fragmention, and the
dictionary will not be defragged.
## Solution
* Add defragmentation of the per-slot dictionary's own structures, dict
struct and ht_table.
## Other change
* Increase floating point print precision of `frags` and `rss` in debug
logs for defrag
---------
Co-authored-by: Oran Agra <oran@redislabs.com>
Currently, once active defrag starts, we can not adjust
active_defrag_running
downwards. This is because active_defrag_running will be dynamically
compute
based on the fragmentation, we think we should not lower the effort when
the
fragmentation drops.
However, we need to note that active_defrag_running will also be
dynamically
computed based on configurations. In this case, we are not respecting
cycle-min
or cycle-max. Some people may realize halfway through that defrag
consumes a
lot and want to adjust it.
Previously we could only turn off activedefrag and then turn it on again
to
adjust active_defrag_running downwards. So in this PR, when a active
defrag
configuration change is made, we will re-compute it.
These configuration items are:
- active-defrag-cycle-min
- active-defrag-cycle-max
- active-defrag-threshold-upper
Since now a DB in cluster mode is divided into 16384 dicts, here
we directly check kvstoreDictSize instead of kvstoreSize, which
may have a higher probability that we can save the lookup.
The other change is a cleanup, obviously kvstoreGetHash should be
applied to the db->expires dicts.
# Description
Gather most of the scattered `redisDb`-related code from the per-slot
dict PR (#11695) and turn it to a new data structure, `kvstore`. i.e.
it's a class that represents an array of dictionaries.
# Motivation
The main motivation is code cleanliness, the idea of using an array of
dictionaries is very well-suited to becoming a self-contained data
structure.
This allowed cleaning some ugly code, among others: loops that run twice
on the main dict and expires dict, and duplicate code for allocating and
releasing this data structure.
# Notes
1. This PR reverts the part of https://github.com/redis/redis/pull/12848
where the `rehashing` list is global (handling rehashing `dict`s is
under the responsibility of `kvstore`, and should not be managed by the
server)
2. This PR also replaces the type of `server.pubsubshard_channels` from
`dict**` to `kvstore` (original PR:
https://github.com/redis/redis/pull/12804). After that was done,
server.pubsub_channels was also chosen to be a `kvstore` (with only one
`dict`, which seems odd) just to make the code cleaner by making it the
same type as `server.pubsubshard_channels`, see
`pubsubtype.serverPubSubChannels`
3. the keys and expires kvstores are currenlty configured to allocate
the individual dicts only when the first key is added (unlike before, in
which they allocated them in advance), but they won't release them when
the last key is deleted.
Worth mentioning that due to the recent change the reply of DEBUG
HTSTATS changed, in case no keys were ever added to the db.
before:
```
127.0.0.1:6379> DEBUG htstats 9
[Dictionary HT]
Hash table 0 stats (main hash table):
No stats available for empty dictionaries
[Expires HT]
Hash table 0 stats (main hash table):
No stats available for empty dictionaries
```
after:
```
127.0.0.1:6379> DEBUG htstats 9
[Dictionary HT]
[Expires HT]
```
Fixing issues described in #12672, started after #11695
Related to #12674
Fixes the `defrag didn't stop' issue.
In some cases of how the keys were stored in memory
defrag_later_item_in_progress was not getting reset once we finish
defragging the later items and we move to the next slot. This stopped
the scan to happen in the later slots and did not get
This is an implementation of https://github.com/redis/redis/issues/10589 that eliminates 16 bytes per entry in cluster mode, that are currently used to create a linked list between entries in the same slot. Main idea is splitting main dictionary into 16k smaller dictionaries (one per slot), so we can perform all slot specific operations, such as iteration, without any additional info in the `dictEntry`. For Redis cluster, the expectation is that there will be a larger number of keys, so the fixed overhead of 16k dictionaries will be The expire dictionary is also split up so that each slot is logically decoupled, so that in subsequent revisions we will be able to atomically flush a slot of data.
## Important changes
* Incremental rehashing - one big change here is that it's not one, but rather up to 16k dictionaries that can be rehashing at the same time, in order to keep track of them, we introduce a separate queue for dictionaries that are rehashing. Also instead of rehashing a single dictionary, cron job will now try to rehash as many as it can in 1ms.
* getRandomKey - now needs to not only select a random key, from the random bucket, but also needs to select a random dictionary. Fairness is a major concern here, as it's possible that keys can be unevenly distributed across the slots. In order to address this search we introduced binary index tree). With that data structure we are able to efficiently find a random slot using binary search in O(log^2(slot count)) time.
* Iteration efficiency - when iterating dictionary with a lot of empty slots, we want to skip them efficiently. We can do this using same binary index that is used for random key selection, this index allows us to find a slot for a specific key index. For example if there are 10 keys in the slot 0, then we can quickly find a slot that contains 11th key using binary search on top of the binary index tree.
* scan API - in order to perform a scan across the entire DB, the cursor now needs to not only save position within the dictionary but also the slot id. In this change we append slot id into LSB of the cursor so it can be passed around between client and the server. This has interesting side effect, now you'll be able to start scanning specific slot by simply providing slot id as a cursor value. The plan is to not document this as defined behavior, however. It's also worth nothing the SCAN API is now technically incompatible with previous versions, although practically we don't believe it's an issue.
* Checksum calculation optimizations - During command execution, we know that all of the keys are from the same slot (outside of a few notable exceptions such as cross slot scripts and modules). We don't want to compute the checksum multiple multiple times, hence we are relying on cached slot id in the client during the command executions. All operations that access random keys, either should pass in the known slot or recompute the slot.
* Slot info in RDB - in order to resize individual dictionaries correctly, while loading RDB, it's not enough to know total number of keys (of course we could approximate number of keys per slot, but it won't be precise). To address this issue, we've added additional metadata into RDB that contains number of keys in each slot, which can be used as a hint during loading.
* DB size - besides `DBSIZE` API, we need to know size of the DB in many places want, in order to avoid scanning all dictionaries and summing up their sizes in a loop, we've introduced a new field into `redisDb` that keeps track of `key_count`. This way we can keep DBSIZE operation O(1). This is also kept for O(1) expires computation as well.
## Performance
This change improves SET performance in cluster mode by ~5%, most of the gains come from us not having to maintain linked lists for keys in slot, non-cluster mode has same performance. For workloads that rely on evictions, the performance is similar because of the extra overhead for finding keys to evict.
RDB loading performance is slightly reduced, as the slot of each key needs to be computed during the load.
## Interface changes
* Removed `overhead.hashtable.slot-to-keys` to `MEMORY STATS`
* Scan API will now require 64 bits to store the cursor, even on 32 bit systems, as the slot information will be stored.
* New RDB version to support the new op code for SLOT information.
---------
Co-authored-by: Vitaly Arbuzov <arvit@amazon.com>
Co-authored-by: Harkrishn Patro <harkrisp@amazon.com>
Co-authored-by: Roshan Khatri <rvkhatri@amazon.com>
Co-authored-by: Madelyn Olson <madelyneolson@gmail.com>
Co-authored-by: Oran Agra <oran@redislabs.com>
We use the C standard assert() in various places in the codebase, which requires NDEBUG to be undefined. We introduced the redisassert.h file in order to allow low level files to access the assert that maps to serverPanic, but this was only applied tactically and is not available broadly.
This PR removes all usage of the standard library asserts and replaces them with an assert that maps to serverPanic. It makes us immune to accidentally setting the NDEBUG flag preventing assertions. I also marked marked the server asserts as "likely" to not execute. I spot checked various points in the code, and it didn't change the code layout on my x86 mac, but it is more consistent with redisassert.h and seems more correct overall.
Technically declaring a prototype with an empty declaration has been deprecated since the early days of C, but we never got a warning for it. C2x will apparently be introducing a breaking change if you are using this type of declarator, so Clang 15 has started issuing a warning with -pedantic. Although not apparently a problem for any of the compiler we build on, if feels like the right thing is to properly adhere to the C standard and use (void).
If a dict has only keys, and no use of values, then a key can be stored directly in a
dict's hashtable. The key replaces the dictEntry. To distinguish between a key and
a dictEntry, we only use this optimization if the key is odd, i.e. if the key has the least
significant bit set. This is true for sds strings, since the sds header is always an odd
number of bytes.
Dict entries are used as a fallback when there is a hash collision. A special dict entry
without a value (only key and next) is used so we save one word in this case too.
This saves 24 bytes per set element for larges sets, and also gains some speed improvement
as a side effect (less allocations and cache misses).
A quick test adding 1M elements to a set using the command below resulted in memory
usage of 28.83M, compared to 46.29M on unstable.
That's 18 bytes per set element on average.
eval 'for i=1,1000000,1 do redis.call("sadd", "myset", "x"..i) end' 0
Other changes:
Allocations are ensured to have at least 8 bits alignment on all systems. This affects 32-bit
builds compiled without HAVE_MALLOC_SIZE (not jemalloc or glibc) in which Redis
stores the size of each allocation, after this change in 8 bytes instead of previously 4 bytes
per allocation. This is done so we can reliably use the 3 least significant bits in a pointer to
encode stuff.
This change deletes the dictGetNext and dictGetNextRef functions, so the
dict API doesn't expose the next field at all.
The bucket function in dictScan is deleted. A separate dictScanDefrag function
is added which takes a defrag alloc function to defrag-reallocate the dict entries.
"Dirty" code accessing the dict internals in active defrag is removed.
An 'afterReplaceEntry' is added to dictType, which allows the dict user
to keep the dictEntry metadata up to date after reallocation/defrag/move.
Additionally, for updating the cluster slot-to-key mapping, after a dictEntry
has been reallocated, we need to know which db a dict belongs to, so we store
a pointer to the db in a new metadata section in the dict struct, which is
a new mechanism similar to dictEntry metadata. This adds some complexity but
provides better isolation.
Improve memory efficiency of list keys
## Description of the feature
The new listpack encoding uses the old `list-max-listpack-size` config
to perform the conversion, which we can think it of as a node inside a
quicklist, but without 80 bytes overhead (internal fragmentation included)
of quicklist and quicklistNode structs.
For example, a list key with 5 items of 10 chars each, now takes 128 bytes
instead of 208 it used to take.
## Conversion rules
* Convert listpack to quicklist
When the listpack length or size reaches the `list-max-listpack-size` limit,
it will be converted to a quicklist.
* Convert quicklist to listpack
When a quicklist has only one node, and its length or size is reduced to half
of the `list-max-listpack-size` limit, it will be converted to a listpack.
This is done to avoid frequent conversions when we add or remove at the bounding size or length.
## Interface changes
1. add list entry param to listTypeSetIteratorDirection
When list encoding is listpack, `listTypeIterator->lpi` points to the next entry of current entry,
so when changing the direction, we need to use the current node (listTypeEntry->p) to
update `listTypeIterator->lpi` to the next node in the reverse direction.
## Benchmark
### Listpack VS Quicklist with one node
* LPUSH - roughly 0.3% improvement
* LRANGE - roughly 13% improvement
### Both are quicklist
* LRANGE - roughly 3% improvement
* LRANGE without pipeline - roughly 3% improvement
From the benchmark, as we can see from the results
1. When list is quicklist encoding, LRANGE improves performance by <5%.
2. When list is listpack encoding, LRANGE improves performance by ~13%,
the main enhancement is brought by `addListListpackRangeReply()`.
## Memory usage
1M lists(key:0~key:1000000) with 5 items of 10 chars ("hellohello") each.
shows memory usage down by 35.49%, from 214MB to 138MB.
## Note
1. Add conversion callback to support doing some work before conversion
Since the quicklist iterator decompresses the current node when it is released, we can
no longer decompress the quicklist after we convert the list.
Small sets with not only integer elements are listpack encoded, by default
up to 128 elements, max 64 bytes per element, new config `set-max-listpack-entries`
and `set-max-listpack-value`. This saves memory for small sets compared to using a hashtable.
Sets with only integers, even very small sets, are still intset encoded (up to 1G
limit, etc.). Larger sets are hashtable encoded.
This PR increments the RDB version, and has an effect on OBJECT ENCODING
Possible conversions when elements are added:
intset -> listpack
listpack -> hashtable
intset -> hashtable
Note: No conversion happens when elements are deleted. If all elements are
deleted and then added again, the set is deleted and recreated, thus implicitly
converted to a smaller encoding.
This includes two fixes:
* We forgot to count non-key reallocs in defragmentation stats.
* Fix the script defrag tests so to make dict entries less signigicant in fragmentation by making the scripts larger.
This assures active defrage will complete and reach desired results.
Some inherent fragmentation might exists in dict entries which we need to ignore.
This lead to occasional CI failures.
Remove scripts defragger since it was broken since #10126 (released in 7.0 RC1).
would crash the server if defragger starts in a server that contains eval scripts.
In #10126 the global `lua_script` dict became a dict to a custom `luaScript` struct with an internal `robj`
in it instead of a generic `sds` -> `robj` dict. This means we need custom code to defrag it and since scripts
should never really cause much fragmentation it makes more sense to simply remove the defrag code for scripts.
# Background
The main goal of this PR is to remove relevant logics on Lua script verbatim replication,
only keeping effects replication logic, which has been set as default since Redis 5.0.
As a result, Lua in Redis 7.0 would be acting the same as Redis 6.0 with default
configuration from users' point of view.
There are lots of reasons to remove verbatim replication.
Antirez has listed some of the benefits in Issue #5292:
>1. No longer need to explain to users side effects into scripts.
They can do whatever they want.
>2. No need for a cache about scripts that we sent or not to the slaves.
>3. No need to sort the output of certain commands inside scripts
(SMEMBERS and others): this both simplifies and gains speed.
>4. No need to store scripts inside the RDB file in order to startup correctly.
>5. No problems about evicting keys during the script execution.
When looking back at Redis 5.0, antirez and core team decided to set the config
`lua-replicate-commands yes` by default instead of removing verbatim replication
directly, in case some bad situations happened. 3 years later now before Redis 7.0,
it's time to remove it formally.
# Changes
- configuration for lua-replicate-commands removed
- created config file stub for backward compatibility
- Replication script cache removed
- this is useless under script effects replication
- relevant statistics also removed
- script persistence in RDB files is also removed
- Propagation of SCRIPT LOAD and SCRIPT FLUSH to replica / AOF removed
- Deterministic execution logic in scripts removed (i.e. don't run write commands
after random ones, and sorting output of commands with random order)
- the flags indicating which commands have non-deterministic results are kept as hints to clients.
- `redis.replicate_commands()` & `redis.set_repl()` changed
- now `redis.replicate_commands()` does nothing and return an 1
- ...and then `redis.set_repl()` can be issued before `redis.replicate_commands()` now
- Relevant TCL cases adjusted
- DEBUG lua-always-replicate-commands removed
# Other changes
- Fix a recent bug comparing CLIENT_ID_AOF to original_client->flags instead of id. (introduced in #9780)
Co-authored-by: Oran Agra <oran@redislabs.com>
The following variable was renamed:
1. lua_caller -> script_caller
2. lua_time_limit -> script_time_limit
3. lua_timedout -> script_timedout
4. lua_oom -> script_oom
5. lua_disable_deny_script -> script_disable_deny_script
6. in_eval -> in_script
The following variables was moved to lctx under eval.c
1. lua
2. lua_client
3. lua_cur_script
4. lua_scripts
5. lua_scripts_mem
6. lua_replicate_commands
7. lua_write_dirty
8. lua_random_dirty
9. lua_multi_emitted
10. lua_repl
11. lua_kill
12. lua_time_start
13. lua_time_snapshot
This commit is in a low risk of introducing any issues and it
is just moving varibales around and not changing any logic.
For diskless replication in swapdb mode, considering we already spend replica memory
having a backup of current db to restore in case of failure, we can have the following benefits
by instead swapping database only in case we succeeded in transferring db from master:
- Avoid `LOADING` response during failed and successful synchronization for cases where the
replica is already up and running with data.
- Faster total time of diskless replication, because now we're moving from Transfer + Flush + Load
time to Transfer + Load only. Flushing the tempDb is done asynchronously after swapping.
- This could be implemented also for disk replication with similar benefits if consumers are willing
to spend the extra memory usage.
General notes:
- The concept of `backupDb` becomes `tempDb` for clarity.
- Async loading mode will only kick in if the replica is syncing from a master that has the same
repl-id the one it had before. i.e. the data it's getting belongs to a different time of the same timeline.
- New property in INFO: `async_loading` to differentiate from the blocking loading
- Slot to Key mapping is now a field of `redisDb` as it's more natural to access it from both server.db
and the tempDb that is passed around.
- Because this is affecting replicas only, we assume that if they are not readonly and write commands
during replication, they are lost after SYNC same way as before, but we're still denying CONFIG SET
here anyways to avoid complications.
Considerations for review:
- We have many cases where server.loading flag is used and even though I tried my best, there may
be cases where async_loading should be checked as well and cases where it shouldn't (would require
very good understanding of whole code)
- Several places that had different behavior depending on the loading flag where actually meant to just
handle commands coming from the AOF client differently than ones coming from real clients, changed
to check CLIENT_ID_AOF instead.
**Additional for Release Notes**
- Bugfix - server.dirty was not incremented for any kind of diskless replication, as effect it wouldn't
contribute on triggering next database SAVE
- New flag for RM_GetContextFlags module API: REDISMODULE_CTX_FLAGS_ASYNC_LOADING
- Deprecated RedisModuleEvent_ReplBackup. Starting from Redis 7.0, we don't fire this event.
Instead, we have the new RedisModuleEvent_ReplAsyncLoad holding 3 sub-events: STARTED,
ABORTED and COMPLETED.
- New module flag REDISMODULE_OPTIONS_HANDLE_REPL_ASYNC_LOAD for RedisModule_SetModuleOptions
to allow modules to declare they support the diskless replication with async loading (when absent, we fall
back to disk-based loading).
Co-authored-by: Eduardo Semprebon <edus@saxobank.com>
Co-authored-by: Oran Agra <oran@redislabs.com>
Redis lists are stored in quicklist, which is currently a linked list of ziplists.
Ziplists are limited to storing elements no larger than 4GB, so when bigger
items are added they're getting truncated.
This PR changes quicklists so that they're capable of storing large items
in quicklist nodes that are plain string buffers rather than ziplist.
As part of the PR there were few other changes in redis:
1. new DEBUG sub-commands:
- QUICKLIST-PACKED-THRESHOLD - set the threshold of for the node type to
be plan or ziplist. default (1GB)
- QUICKLIST <key> - Shows low level info about the quicklist encoding of <key>
2. rdb format change:
- A new type was added - RDB_TYPE_LIST_QUICKLIST_2 .
- container type (packed / plain) was added to the beginning of the rdb object
(before the actual node list).
3. testing:
- Tests that requires over 100MB will be by default skipped. a new flag was
added to 'runtest' to run the large memory tests (not used by default)
Co-authored-by: sundb <sundbcn@gmail.com>
Co-authored-by: Oran Agra <oran@redislabs.com>
Part two of implementing #8702 (zset), after #8887.
## Description of the feature
Replaced all uses of ziplist with listpack in t_zset, and optimized some of the code to optimize performance.
## Rdb format changes
New `RDB_TYPE_ZSET_LISTPACK` rdb type.
## Rdb loading improvements:
1) Pre-expansion of dict for validation of duplicate data for listpack and ziplist.
2) Simplifying the release of empty key objects when RDB loading.
3) Unify ziplist and listpack data verify methods for zset and hash, and move code to rdb.c.
## Interface changes
1) New `zset-max-listpack-entries` config is an alias for `zset-max-ziplist-entries` (same with `zset-max-listpack-value`).
2) OBJECT ENCODING will return listpack instead of ziplist.
## Listpack improvements:
1) Add `lpDeleteRange` and `lpDeleteRangeWithEntry` functions to delete a range of entries from listpack.
2) Improve the performance of `lpCompare`, converting from string to integer is faster than converting from integer to string.
3) Replace `snprintf` with `ll2string` to improve performance in converting numbers to strings in `lpGet()`.
## Zset improvements:
1) Improve the performance of `zzlFind` method, use `lpFind` instead of `lpCompare` in a loop.
2) Use `lpDeleteRangeWithEntry` instead of `lpDelete` twice to delete a element of zset.
## Tests
1) Add some unittests for `lpDeleteRange` and `lpDeleteRangeWithEntry` function.
2) Add zset RDB loading test.
3) Add benchmark test for `lpCompare` and `ziplsitCompare`.
4) Add empty listpack zset corrupt dump test.
Add two INFO metrics:
```
total_active_defrag_time:12345
current_active_defrag_time:456
```
`current_active_defrag_time` if greater than 0, means how much time has
passed since active defrag started running. If active defrag stops, this metric is reset to 0.
`total_active_defrag_time` means total time the fragmentation
was over the defrag threshold since the server started.
This is a followup PR for #9031
* Enhance dict to support arbitrary metadata carried in dictEntry
Co-authored-by: Viktor Söderqvist <viktor.soderqvist@est.tech>
* Rewrite slot-to-keys mapping to linked lists using dict entry metadata
This is a memory enhancement for Redis Cluster.
The radix tree slots_to_keys (which duplicates all key names prefixed with their
slot number) is replaced with a linked list for each slot. The dict entries of
the same cluster slot form a linked list and the pointers are stored as metadata
in each dict entry of the main DB dict.
This commit also moves the slot-to-key API from db.c to cluster.c.
Co-authored-by: Jim Brunner <brunnerj@amazon.com>
Part one of implementing #8702 (taking hashes first before other types)
## Description of the feature
1. Change ziplist encoded hash objects to listpack encoding.
2. Convert existing ziplists on RDB loading time. an O(n) operation.
## Rdb format changes
1. Add RDB_TYPE_HASH_LISTPACK rdb type.
2. Bump RDB_VERSION to 10
## Interface changes
1. New `hash-max-listpack-entries` config is an alias for `hash-max-ziplist-entries` (same with `hash-max-listpack-value`)
2. OBJECT ENCODING will return `listpack` instead of `ziplist`
## Listpack improvements:
1. Support direct insert, replace integer element (rather than convert back and forth from string)
3. Add more listpack capabilities to match the ziplist ones (like `lpFind`, `lpRandomPairs` and such)
4. Optimize element length fetching, avoid multiple calculations
5. Use inline to avoid function call overhead.
## Tests
1. Add a new test to the RDB load time conversion
2. Adding the listpack unit tests. (based on the one in ziplist.c)
3. Add a few "corrupt payload: fuzzer findings" tests, and slightly modify existing ones.
Co-authored-by: Oran Agra <oran@redislabs.com>
Reduce dict struct memory overhead
on 64bit dict size goes down from jemalloc's 96 byte bin to its 56 byte bin.
summary of changes:
- Remove `privdata` from callbacks and dict creation. (this affects many files, see "Interface change" below).
- Meld `dictht` struct into the `dict` struct to eliminate struct padding. (this affects just dict.c and defrag.c)
- Eliminate the `sizemask` field, can be calculated from size when needed.
- Convert the `size` field into `size_exp` (exponent), utilizes one byte instead of 8.
Interface change: pass dict pointer to dict type call back functions.
This is instead of passing the removed privdata field. In the future if
we'd like to have private data in the callbacks we can extract it from
the dict type. We can extend dictType to include a custom dict struct
allocator and use it to allocate more data at the end of the dict
struct. This data can then be used to store private data later acccessed
by the callbacks.
Create new module type enhanced callbacks: mem_usage2, free_effort2, unlink2, copy2.
These will be given a context point from which the module can obtain the key name and database id.
In addition the digest and defrag context can now be used to obtain the key name and database id.
In activeDefragSdsListAndDict when defrag list key sucess, the key
in dict is also replaced. Then the corresponding dictEntry is
defraged. But the dictEntry will be defraged in next step by defrag
the dict values too.
In activeDefragSdsListAndDict when dict_val_type is DEFRAG_SDS_DICT_VAL_VOID_PTR, it should update de->v.val not ln->value.
Because this code path will never be executed, so this bug never happened.
Add a new set of defrag functions that take a defrag context and allow
defragmenting memory blocks and RedisModuleStrings.
Modules can register a defrag callback which will be invoked when the
defrag process handles globals.
Modules with custom data types can also register a datatype-specific
defrag callback which is invoked for keys that require defragmentation.
The callback and associated functions support both one-step and
multi-step options, depending on the complexity of the key as exposed by
the free_effort callback.
When loading an encoded payload we will at least do a shallow validation to
check that the size that's encoded in the payload matches the size of the
allocation.
This let's us later use this encoded size to make sure the various offsets
inside encoded payload don't reach outside the allocation, if they do, we'll
assert/panic, but at least we won't segfault or smear memory.
We can also do 'deep' validation which runs on all the records of the encoded
payload and validates that they don't contain invalid offsets. This lets us
detect corruptions early and reject a RESTORE command rather than accepting
it and asserting (crashing) later when accessing that payload via some command.
configuration:
- adding ACL flag skip-sanitize-payload
- adding config sanitize-dump-payload [yes/no/clients]
For now, we don't have a good way to ensure MIGRATE in cluster resharding isn't
being slowed down by these sanitation, so i'm setting the default value to `no`,
but later on it should be set to `clients` by default.
changes:
- changing rdbReportError not to `exit` in RESTORE command
- adding a new stat to be able to later check if cluster MIGRATE isn't being
slowed down by sanitation.
Since the dynamic allocations in raxIterator are only used for deep walks, memory
leak due to missing call to raxStop can only happen for rax with key names longer
than 32 bytes.
Out of all the missing calls, the only ones that may lead to a leak are the rax
for consumer groups and consumers, and these were only in AOFRW and rdbSave, which
normally only happen in fork or at shutdown.
There's a rare case which leads to stagnation in the defragger, causing
it to keep scanning the keyspace and do nothing (not moving any
allocation), this happens when all the allocator slabs of a certain bin
have the same % utilization, but the slab from which new allocations are
made have a lower utilization.
this commit fixes it by removing the current slab from the overall
average utilization of the bin, and also eliminate any precision loss in
the utilization calculation and move the decision about the defrag to
reside inside jemalloc.
and also add a test that consistently reproduce this issue.
When active defrag kicks in and finds a big list, it will create a bookmark to
a node so that it is able to resume iteration from that node later.
The quicklist manages that bookmark, and updates it in case that node is deleted.
This will increase memory usage only on lists of over 1000 (see
active-defrag-max-scan-fields) quicklist nodes (1000 ziplists, not 1000 items)
by 16 bytes.
In 32 bit build, this change reduces the maximum effective config of
list-compress-depth and list-max-ziplist-size (from 32767 to 8191)
Reduce default minimum effort, so that when fragmentation is just detected,
the impact on the latency will be minor.
Reduce the default maximum effort, mainly to prevent a case were a sudden
massive deletions, won't trigger an aggressive defrag that will cause latency.
When activedefrag is disabled mid-run, reset the 'running' info field, and
clear the scan cursor, so that when it'll be re-enabled, a new fresh scan will
start.
Clearing the 'running' variable is important since lowering the defragger
tunables mid-scan won't help, the defragger only considers new threshold when
a new scan starts, and during a scan it can only become more aggressive,
(when more severe fragmentation is detected), it'll never go less aggressive.
So by temporarily disabling activedefrag, one can lower th the tunables.
Removing the experimantal warning.
debing.sun