system.batchlog will still have to be used while the cluster is
upgrading from an older version, which doesn't know v2 yet.
Re-add support for replaying v1 batchlogs. The switch to v2 will happen
after the BATCHLOG_V2 cluster feature is enabled.
The only external user -- storage_proxy -- only needs a minor
adjustment: switch between the table names. The rest is handled
transparently by the db/batchlog.hh interface and the batchlog_manager.
Although the value of this item is indeed derived from the write timeout
config, the name doesn't reflect what it is used for. Change it to
reflect it better.
New batchlogs are written to the batchlog_v2 table and replay also uses
the v2 table.
The content of system.batchlog is attempted to be migrated to
system.batchlog_v2 after each start of the batchlog_manager service.
The migration is retried on each replay if it fails. This is reduntant
but simple.
Batchlog cleanup now doesn't involve flushing memtables, the only
remaining user of replica/database.hh is gone, so the include is
dropped.
Return a flag determining whether all the batches were sent successfully in
batchlog_manager::replay_all_failed_batches (batches skipped due to being
too fresh are not counted). Throw in repair_flush_hints_batchlog_handler
if not all batches were replayed, to ensure that repair_time isn't updated.
Add a test of the batchlog manager replay loop applying failed batches
while some replica is down.
The test reproduces an issue where the batchlog manager tries to replay
a failed batch, doesn't get a response from some replica, and becomes
stuck.
It verifies that the batchlog manager can eventually recover from this
situation and continue applying failed batches.
Set a timeout on writes of replayed batches by the batchlog manager.
We want to avoid having infinite timeout for the writes in case it gets
stuck for some unexpected reason.
The timeout is set to be high enough to allow any reasonable write to
complete.
After the specified amount of replays, trigger a cleanup: flush batchlog
table memtables. This allows the cleanup to happen on a configurable
interval, instead of on every batchlog replay attempt, which might be
too much.
Add a flag controlling whether cleanup (memtable flush) will be done
after the replay. This is to allow repair to opt out from cleanup --
when many concurrenty repairs are running, there can be storms of calles
to do_batch_log_replay(), which will be mostly no-op, but they will all
attempt to flush the memtable to clean-up after themselves. This is
unnecessary and introduces latency to repairs, best to leave the cleanup
to the periodic batch-log replay.
these unused includes are identified by clang-include-cleaner.
after auditing the source files, all of the reports have been
confirmed.
please note, since we have `using seastar::shared_ptr` in
`seastarx.h`, this renders `#include <seastar/core/shared_ptr.hh>`
unnecessary if we don't need the full definition of `seastar::shared_ptr`.
so, in this change, all the unused includes are removed. but there are
some headers which are actually used, while still being identified by
this tool. these includes are marked with "IWYU pragma: keep".
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
In b4e66ddf1d (4.0) we added a new batchlog_manager configuration
named delay, but forgot to initialize it in cql_test_env. This somehow
worked, but doesn't with clang 18.
Fix it by initializing to 0 (there isn't a good reason to delay it).
Also provide a default to make it safer.
Closesscylladb/scylladb#18572
... and sanitize the future used on stop.
The loop in question is now started in .start(), but all callers now
construct the manager late enough, so the loop spawning can be moved.
This also calls for renaming the future member of the class and allows
to make it regular, not shared, future.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
The manager will need system ks to get truncation record from, so add it
explicitly. Start-stop sequence no allows that
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Move mutation-related files to a new mutation/ directory. The names
are kept in the global namespace to reduce churn; the names are
unambiguous in any case.
mutation_reader remains in the readers/ module.
mutation_partition_v2.cc was missing from CMakeLists.txt; it's added in this
patch.
This is a step forward towards librarization or modularization of the
source base.
Closes#12788
The .drain() method can be called from several places, each needs to
wait for its completion. Now this is achieved with the help of a gate,
but there's a simpler way
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
After fcb8d040 ("treewide: use Software Package Data Exchange
(SPDX) license identifiers"), many dual-licensed files were
left with empty comments on top. Remove them to avoid visual
noise.
Closes#10562
Instead of lengthy blurbs, switch to single-line, machine-readable
standardized (https://spdx.dev) license identifiers. The Linux kernel
switched long ago, so there is strong precedent.
Three cases are handled: AGPL-only, Apache-only, and dual licensed.
For the latter case, I chose (AGPL-3.0-or-later and Apache-2.0),
reasoning that our changes are extensive enough to apply our license.
The changes we applied mechanically with a script, except to
licenses/README.md.
Closes#9937
The header file <gm/inet_address.hh> is included, directly or
indirectly, from 291 source files in Scylla. It is hard to reduce this
number because Scylla relies heavily on IP addresses as keys to
different things. So it is important that this header file be fast to
include. Unfortunately it wasn't... ClangBuildAnalyzer measurements
showed that each inclusion of this header file added a whopping 2 seconds
(in dev build mode) to the build. A total of 600 CPU seconds - 10 CPU
minutes - were spent just on this header file. It was actually worse
because the build also spent additional time on template instantiation
(more on this below).
So in this patch we:
1. Remove some unnecessary stuff from gms/inet_address.hh, and avoid
including it in one place that doesn't need it. This is just
cosmetic, and doesn't significantly speed up the build.
2. Move the to_sstring() implementation for the .hh to .cc. This saves
a lot of time on template instantiations - previously every source
file instantiated this to_sstring(), which was slow (that "format"
thing is slow).
3. Do not include <seastar/net/ip.hh> which is a huge file including
half the world. All we need from it is the type "ipv4_address",
so instead include just the new <seastar/net/ipv4_address.hh>.
This change brings most of the performance improvement.
So source files forgot to include various Seastar header files
because the includes-everything ip.hh did it - so we need to add
these missing includes in this patch.
After this patch, ClangBuildAnalyzer's reports that the cost of
inclusion of <gms/inet_address.hh> is down from 2 seconds to 0.326
seconds. Additionally the format<inet_address> template instantiation
291 times - about half a second each - is also gone.
All in all, this patch should reduce around 10 CPU minutes from the build.
Refs #1
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
And rename to get_batchlog_mutation_for while at it,
as it's about the batchlog, not batch_log.
This resolves a circular dependency between the
batchlog_manager and the storage_proxy that required
it in the case.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
There's nothing in this function that actually requries
the batchlog manager instance.
It uses a random number engine that's moved along with it
to class gossiper.
This resolves a circular dependency between the
batchlog_manager and storage_proxy.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
As a prerequisite to globalizing the batchlog_manager,
allow setting a global pointer to it and instantiate
the sharded<db::batchlog_manager> on the main/cql_test_env
stack.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
drain() aborts the replay loop fiber
and returns its future.
It's grabbing _gate so stop() will wait on it.
The intention is to call stop_replay_loop from
storage_service::decommission and do_drain rather
than stop, so we can stop the batchlog manager once,
using a deferred action in main.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Harden start/stop by using an abort_source to abort from
the replay loop.
Extract the loop into batchlog_replay_loop() coroutine,
with the _stop abourt source as a stop condition,
plus use it for sleep_abortable to be able to promptly
stop while sleeping.
start() stores batchlog_replay_loop's future in a newly added
_started member, which is waited on in stop() to synchronize
with the start process at any stage.
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
The write paths in storage_proxy pass replica sets as
std::unordered_set<gms::inet_address>. This is a complex type, with
N+1 allocations for N members, so we change it to a small_vector (via
inet_address_vector_replica_set) which requires just one allocation, and
even zero when up to three replicas are used.
This change is more nuanced than the corresponding change to the read path
abe3d7d7 ("Merge 'storage_proxy: use small_vector for vectors of
inet_address' from Avi Kivity"), for two reasons:
- there is a quadratic algorithm in
abstract_write_response_handler::response(): it searches for a replica
and erases it. Since this happens for every replica, it happens N^2/2
times.
- replica sets for writes always include all datacenters, while reads
usually involve just one datacenter.
So, a write to a keyspace that has 5 datacenters will invoke 15*(15-1)/2
=105 compares.
We could remove this by sending the index of the replica in the replica
set to the replica and ask it to include the index in the response, but
I think that this is unnecessary. Those 105 compares need to be only
105/15 = 7 times cheaper than the corresponding unordered_set operation,
which they surely will. Handling a response after a cross-datacenter round
trip surely involves L3 cache misses, and a small_vector reduces these
to a minimum compared to an unordered_set with its bucket table, linked
list walking and managent, and table rehashing.
Tests using perf_simple_query --write --smp 1 --operations-per-shard 1000000
--task-quota-ms show two allocations removed (as expected) and a nice
reduction in instructions executed.
before: median 204842.54 tps ( 54.2 allocs/op, 13.2 tasks/op, 49890 insns/op)
after: median 206077.65 tps ( 52.2 allocs/op, 13.2 tasks/op, 49138 insns/op)
Closes#8847
Eliminate not used includes and replace some more includes
with forward declarations where appropriate.
Signed-off-by: Pavel Solodovnikov <pa.solodovnikov@scylladb.com>
Extract configuration into a new struct batchlog_manager_config and have the
callers populate it using db::config. This reduces dependencies on global objects.
std::random_device() uses the relatively slow /dev/urandom, and we rarely if
ever intend to use it directly - we normally want to use it to seed a faster
random_engine (a pseudo-random number generator).
In many places in the code, we first created a random_device variable, and then
using it created a random_engine variable. However, this practice created the
risk of a programmer accidentally using the random_device object, instead of the
random_engine object, because both have the same API; This hurts performance.
This risk materialized in just two places in the code, utils/uuid.cc and
gms/gossiper.cc. A patch for to uuid.cc was sent previously by Pawel and is
not included in this patch, and the fix for gossiper.{cc,hh} is included here.
To avoid risking the same mistake in the future, this patch switches across the
code to an idiom where the random_device object is not *named*, so cannot be
accidentally used. We use the following idiom:
std::default_random_engine _engine{std::random_device{}()};
Here std::random_device{}() creates the random device (/dev/urandom) and pulls
a random integer from it. It then uses this seed to create the random_engine
(the pseudo-random number generator). The std::random_device{} object is
temporary and unnamed, and cannot be unintentionally used directly.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Message-Id: <20180726154958.4405-1-nyh@scylladb.com>
* seastar f14d2a3...7a49ae5 (8):
> sharded: improve support for cooperating sharded<> services
> sharded: support for peer services
> semaphore: add a version of with_semaphore that takes a duration timeout
> scripts: perftune.py: fix the CPU mask generation for more than 64 CPUs
> Revert "future-utils: make when_all() (vector variant) exception safe"
> Revert "future-utils: fix gross compilation errors in when_all()"
> future-utils: fix gross compilation errors in when_all()
> future-utils: make when_all() (vector variant) exception safe
Includes change to batchlog_manager constructor to adapt it to
seastar::sharded::start() change.
During decommission, the storage_service::unbootstrap() needs to
initiate a batchlog replay operation. To sync the replay operation
initiated by the timer in batchlog_manager and storage_service, a
semaphore is introduced. To simplify the semaphore locking, the
management code now always runs on shard zero, but the real work is
distruted to all shards.
