Current native restore does not support primary_replica_only, it is
hard-coded disabled and this may lead to data amplification issues.
This patch extends the restore REST API to accept a
primary_replica_only parameter and propagates it to
sstables_loader so it gets correctly passed to
load_and_stream.
Fixes#26584
Signed-off-by: Robert Bindar <robert.bindar@scylladb.com>
Signed-off-by: Robert Bindar <robert.bindar@scylladb.com>
Load-and-stream is broken when running concurrently to the
finalization step of tablet split.
Consider this:
1) split starts
2) split finalization executes barrier and succeed
3) load-and-stream runs now, starts writing sstable (pre-split)
4) split finalization publishes changes to tablet metadata
5) load-and-stream finishes writing sstable
6) sstable cannot be loaded since it spans two tablets
two possible fixes (maybe both):
1) load-and-stream awaits for topology to quiesce
2) perform split compaction on sstable that spans both sibling tablets
This patch implements #1. By awaiting for topology to quiesce,
we guarantee that load-and-stream only starts when there's no
chance coordinator is handling some topology operation like
split finalization.
Fixes#26455.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Change return type of `check_needs_view_update_path()`. Instead of
retrning bool which tells whether to use staging directory (and register
to `view_update_generator`) or use normal directory.
Now the function returns enum with possible values:
- `normal_directory` - use normal directory for the sstable
- `staging_directly_to_generator` - use staging directory and register
to `view_update_generator`
- `staging_managed_by_vbc` - use staging directory but don't register it
to `view_update_generator` but create view building tasks for
later
The third option is new, it's used when the table has any view which is
in building process currrently. In this case, registering it to `view_update_generator`
prematurely may lead to base-view inconsistency
(for example when a replica is in a pending state).
This patch adds the new option in nodetool, patches the
load_new_ss_tables REST request with a new parameter and
skips the reshape step in refresh if this flag is passed.
Signed-off-by: Robert Bindar <robert.bindar@scylladb.com>
Closesscylladb/scylladb#24409Fixes: #24365
Just put the boolean into the callstack between API and distributed
loader to reduce the churn in the next patches. No functional changes,
flag is false and unused.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Previously, download_task_impl's destructor would destroy per-shard progress
elements on whatever shard the task was destroyed on. In multi-shard
environments, this caused "shared_ptr accessed on non-owner cpu" errors when
attempting to free memory allocated on a different shard.
Fix by:
- Convert progress_per_shard into a sharded service
- Stop the service on owner shards during cleanup using coroutines
- Add operator+= to stream_progress to leverage seastar's built-in adder
instead of a custom adder struct
Alternative approaches considered:
1. Using foreign_ptr: Rejected as it would require interface changes
that complicate stream delegation. foreign_ptr manages the underlying
pointee with another smart pointer but does not expose the smart
pointer instance in its APIs, making it impossible to use
shared_ptr<stream_progress> in the interface.
2. Using vector<stream_progress>: Rejected for similar interface
compatibility reasons.
This solution maintains the existing interfaces while ensuring proper
cross-shard cleanup.
Fixesscylladb/scylladb#22759
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
We restore a snapshot of table by streaming the sstables of
the given snapshot of the table using
`sstable_streamer::stream_sstable_mutations()` in batches. This function
reads mutations from a set of sstables, and streams them to the target
nodes. Due to the limit of this function, we are not able to track the
progress in bytes.
Previously, progress tracking used individual sstables as units, which caused
inaccuracies with tablet-distributed tables, where:
- An sstable spanning multiple tablets could be counted multiple times
- Progress reporting could become misleading (e.g., showing "40" progress
for a table with 10 sstables)
This change introduces a more robust progress tracking method:
- Use "batch" as the unit of progress instead of individual sstables.
Each batch represents a tablet when restoring a table snapshot if
the tablet being restored is distributed with tablets. When it comes
to tables distributed with vnode, each batch represents an sstable.
- Stream sstables for each tablet separately, handling both partially and
fully contained sstables
- Calculate progress based on the total number of sstables being streamed
- Skip tablet IDs with no owned tokens
For vnode-distributed tables, the number of "batches" directly corresponds
to the number of sstables, ensuring:
- Consistent progress reporting across different table distribution models
- Simplified implementation
- Accurate representation of restore progress
The new approach provides a more reliable and uniform method of tracking
restoration progress across different table distribution strategies.
Also, Corrected the use of `_sstables.size()` in
`sstable_streamer::stream_sstables()`. It addressed a review comment
from Pavel that was inadvertently overlooked during previous rebasing
the commit of 5ab4932f34.
Fixesscylladb/scylladb#21816
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
Closesscylladb/scylladb#21841
Semi-mechanical change that adds newly introduced "scope" parameter to
all the functions between API methods and the low-level streamer object.
No real functional changes. API methods set it to "all" to keep existing
behavior.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Currently load-and-stream sends mutations to whatever node is considered
to be a "replica" for it. One exception is the "primary-replica-only"
flag that can be requested by the user.
This patch introduces a "scope" parameter that limits streaming part in
where it can stream the data to with 4 options:
- all -- current way of doing things, stream to wherever needed
- dc -- only stream to nodes that live in the same datacenter
- rack -- only stream to nodes that live in the same rack
- node -- only "stream" to current node
It's not yet configurable and streamer object initializes itself with
"all" mode. Will be changed later.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Previously, the progress of download_task_impl launched by the "restore" API
was not tracked. Since restore operations can involve large data transfers,
this makes it difficult for users to monitor progress.
The restore process happens in two sequential steps:
1. Open specified SSTables from object storage
2. Download and stream mutation fragments from the opened SSTables to
mapped destinations
While both steps contribute to overall progress, they use different units
of measurement, making a unified progress metric challenging. Because
the load-and-stream step (step 2) is the largest time-consuming part of the
restore. This change implements progress tracking for this step as an
initial improvement to provide users with partial visibility into the
restore operation.
Fixesscylladb/scylladb#21427
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
The restore-from-s3 task uses load-and-stream internally which, in turn, unlinks loaded sstables on success. That's not what user expects when it restores from backup, objects should remain in bucket afterwards.
Closesscylladb/scylladb#20947
* github.com:scylladb/scylladb:
test: Add check that restored-from objects are not removed
sstables_loader: Dont unlink sstables when restoring from S3
sstables_loader: Make primary_replica_only bool_class RAII field
When load_and_stream() completes, all sstables that were loaded (and
streamed) are unlinked. This is wrong for the restore-from-s3 task, as
removing objects from backup storage is not what user expects.
Fix it by adding a boolean to streamer class, and set it to false (well,
bool_class<>::no) for restore task.
fixes: #20938
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
before this change, we enumerate the sstables tracked by the
system.sstables table, and restore them when serving
requests to "storage_service/restore" API. this works fine with
"storage_service/backup" API. but this "restore" API cannot be
used as a drop-in replacement of the rclone based API currently
used by scylla-manager.
in order to fill the gap, in this change:
* add the "prefix" parameter for specifying the shared prefix of
sstables
* add the "sstables" parameter for specifying the list of TOC
components of sstables
* remove the "snapshot" parameter, as we don't encode the prefix
on scylla's end anymore.
* make the "table" parameter mandatory.
Fixesscylladb/scylladb#20461
Signed-off-by: Kefu Chai <kefu.chai@scylladb.com>
The method starts a task that uses sstables_loader load-and-stream
functionality to bring new sstables into the cluster. The existing
load-and-stream picks up sstables from upload/ directory, the newly
introduced task collects them from S3 bucket and given prefix (that
correspond to the path where backup API method put them).
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
The storage_manager maintains set of clients to configured object
storage(s). The sstables loader is going to spawn tasks that will talk
to to those storages, thus it needs the storage manager to get the
clients clients from.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
This service is going to start tasks managed by task manager. For that,
it should have its module set up and registered.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
Schema related files are moved there. This excludes schema files that
also interact with mutations, because the mutation module depends on
the schema. Those files will have to go into a separate module.
Closes#12858
Define table_id as a distinct utils::tagged_uuid modeled after raft
tagged_id, so it can be differentiated from other uuid-class types,
in particular from table_schema_version.
Fixes#11207
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
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
Move replica-oriented classes to the replica namespace. The main
classes moved are ::database, ::keyspace, and ::table, but a few
ancillary classes are also moved. There are certainly classes that
should be moved but aren't (like distributed_loader) but we have
to start somewhere.
References are adjusted treewide. In many cases, it is obvious that
a call site should not access the replica (but the data_dictionary
instead), but that is left for separate work.
scylla-gdb.py is adjusted to look for both the new and old names.
The code was moved in the relevant .cc file by previous patch, now
make it sit in the relevant class. One "significant" change is that
the messaging service is available by local reference already, not
by the sharded one. Other dependencies are already satisfied by the
patch that introduced the sstables_loader class.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
It's a sharded service that will be responsible for loading
sstables via the respective REST API (the endpoint in question
is in turn handling the nodetool refresh command). This patch
adds the loader, equips with the needed dependencies and
starts/stops one from main. Next patches will move the loader
code from storage_service into this new one. The list of
dependencies that are introduced in this patch is exactly
what's needed by the mentioned code move.
Signed-off-by: Pavel Emelyanov <xemul@scylladb.com>
