Add `--incremental-snapshots` flag to enable incremental snapshots.
This will allow setting `--full-snapshot-interval-slots` and
`--incremental-snapshot-interval-slots`.
Also added `--maximum-incremental-snapshots-to-retain`.
Co-authored-by: Michael Vines <mvines@gmail.com>
This is the 2nd installment for the AccountsDb replication.
Summary of Changes
The basic google protocol buffer protocol for replicating updated slots and accounts. tonic/tokio is used for transporting the messages.
The basic framework of the client and server for replicating slots and accounts -- the persisting of accounts in the replica-side will be done at the next PR -- right now -- the accounts are streamed to the replica-node and dumped. Replication for information about Bank is also not done in this PR -- to be addressed in the next PR to limit the change size.
Functionality used by both the client and server side are encapsulated in the replica-lib crate.
There is no impact to the existing validator by default.
Tests:
Observe the confirmed slots replicated to the replica-node.
Observe the accounts for the confirmed slot are received at the replica-node side.
AccountsBackgroundService now knows about incremental snapshots. It is
now also in charge of deciding if an AccountsPackage is destined to be a
SnapshotPackage or not (or just used by AccountsHashVerifier).
!!! New behavior changes !!!
Taking snapshots (both bank and archive) **MUST** succeed.
This is required because of how the last full snapshot slot is
calculated, which is used by AccountsBackgroundService when calling
`clean_accounts()`.
File system calls are now unwrapped and will result in a crash. As Trent told me:
>Well I think if a snapshot fails due to some IO error, it's very likely that the operator is going to have to intervene before it works. We should exit error in this case, otherwise the validator might happily spin for several more hours, never successfully writing a complete snapshot, before something else brings it down. This would leave the validator's last local snapshot many more slots behind than it would be had we exited outright and potentially force the operator to abandon ledger continuity in favor of a quick catchup
Other errors will set the `exit` flag to `true`, and the node will gracefully shutdown.
Fixes#19167Fixes#19168
#### Problem
Snapshot names are overloaded, and there are multiple terms that mean the same thing. This is confusing. Here's a list of ones in the codebase that I've found:
```
- snapshot_dir
- snapshots_dir
- snapshot_path
- snapshot_output_dir
- snapshot_package_output_path
- snapshot_archives_dir
```
#### Summary of Changes
For all the ones that are about the directory where snapshot archives are stored, ensure they are `snapshot_archives_dir`. For the ones about the (bank) snapshots directory, set to `bank_snapshots_dir`.
Co-authored-by: Michael Vines <mvines@gmail.com>
Shreds recovered from erasure codes have not been received from turbine
and have not been retransmitted to other nodes downstream. This results
in more repairs across the cluster which is slower.
This commit channels through recovered shreds to retransmit stage in
order to further broadcast the shreds to downstream nodes in the tree.
Working towards sending shreds (instead of packets) to retransmit stage
so that shreds recovered from erasure codes are as well retransmitted.
Following commit will add these metrics back to window-service, earlier
in the pipeline.
Cross cluster gossip contamination is causing cluster-slots hash map to
contain a lot of bogus values and consume too much memory:
https://github.com/solana-labs/solana/issues/17789
If a node is using the same identity key across clusters, then these
erroneous values might not be filtered out by shred-versions check,
because one of the variants of the contact-info will have matching
shred-version:
https://github.com/solana-labs/solana/issues/17789#issuecomment-896304969
The cluster-slots hash-map is bounded and trimmed at the lower end by
the current root. This commit also discards slots epochs ahead of the
root.
Renaming these types to better communicate their usages, which will
further diverge as incremental snapshot support is added.
With the new names, AccountsPacakge now refers to the type between
AccountsBackgroundProcess and AccountsHashVerifier, and SnapshotPackage
refers to the type between AccountsHashVerifier and
SnapshotPackagerService.
* added realtime cost checking logic to reject block that would exceed max limit:
- defines max limits at block_cost_limits.rs
- right after each bath's execution, accumulate its cost and check again
limit, return error if limit is exceeded
* update abi that changed due to adding additional TransactionError
* To avoid counting stats mltiple times, only accumulate execute-timing when a bank is completed
* gate it by a feature
* move cost const def into block_cost_limits.rs
* redefine the cost for signature and account access, removed signer part as it is not well defined for now
* check if per_program_timings of execute_timings before sending
AccountsHashVerifier will need access to both the full and incremental
snapshot archive interval slots config values, which is in the
SnapshotConfig.
Also, cleanup some `Option<>` params and their references.
While reviewing PR #18565, as issue was brought up to refactor some code
around verifying the bank after rebuilding from snapshots. A new
top-level function has been added to get the latest snapshot archives
and load the bank then verify. Additionally, new tests have been
written and existing tests have been updated to use this new function.
Fixes#18973
While resolving the issue, it became clear there was some additional
low-hanging fruit this change enabled. Specifically, the functions
`bank_to_xxx_snapshot_archive()` now return their respective
`SnapshotArchiveInfo`. And on the flip side,
`bank_from_snapshot_archives()` now takes `SnapshotArchiveInfo`s instead
of separate paths and archive formats. This bundling simplifies bank
rebuilding.
bank.get_leader_schedule_epoch(shred_slot)
is one epoch after epoch_schedule.get_epoch(shred_slot).
At epoch boundaries, shred is already one epoch after the root-slot. So
we need epoch-stakes 2 epochs ahead of the root. But the root bank only
has epoch-stakes for one epoch ahead, and as a result looking up epoch
staked-nodes from the root-bank fails.
To be backward compatible with the current master code, this commit
implements a fallback on working-bank if epoch staked-nodes obtained
from the root-bank is none.
If two threads simultaneously call into ClusterNodesCache::get for the
same epoch, and the cache entry is outdated, then both threads recompute
cluster-nodes for the epoch and redundantly overwrite each other.
This commit wraps ClusterNodesCache entries in Arc<Mutex<...>>, so that
when needed only one thread does the computations to update the entry.
* Current caching mechanism does not update cluster-nodes when the epoch
(and so epoch staked nodes) changes:
https://github.com/solana-labs/solana/blob/19bd30262/core/src/broadcast_stage/standard_broadcast_run.rs#L332-L344
* Additionally, the cache update has a concurrency bug in which the
thread which does compare_and_swap may be blocked when it tries to
obtain the write-lock on cache, while other threads will keep running
ahead with the outdated cache (since the atomic timestamp is already
updated).
In the new ClusterNodesCache, entries are keyed by epoch, and so if
epoch changes cluster-nodes will be recalculated. The time-to-live
eviction policy is also encapsulated and rigidly enforced.
The new cluster-nodes cache will:
* ensure cluster-nodes are recalculated if the epoch (and so the epoch
staked nodes) changes.
* encapsulate time-to-live eviction policy.
Cluster nodes are cached keyed by the respective epoch from which stakes
are obtained, and so if epoch changes cluster-nodes will be recomputed.
A time-to-live eviction policy is enforced to refresh entries in case
gossip contact-infos are updated.
This commit also renames `snapshot_interval_slots` to
`full_snapshot_archive_interval_slots`, updates the comments on the
fields, and make appropriate updates where SnapshotConfig is used.
* wip - cost_update_services to log both us and cu for each instruction to determine possible ratio
* replace microsecond with compute_unit as cost unit
This commit adds high-level functions for creating and loading-from
incremental snapshots, plus all low-level functions required to perform
those tasks. This commit **does not** add taking incremental snapshots
as part of a running validator, nor starting up a node with an
incremental snapshot; just laying ground work.
Additionally, `snapshot_utils` and `serde_snapshot` have been
refactored to use a common code paths for the different snapshots.
Also of note, some renaming has happened:
1. Snapshots are now either `full_` or `incremental_` throughout the
codebase. If not specified, the code applies to both.
2. Bank snapshots now are called "bank snapshots"
(before they were called "slot snapshots", "bank snapshots", or
just "snapshots"). The one exception is within `Bank`, where they
are still just "snapshots", because they are already "bank
snapshots".
3. Snapshot archives now have `_archive` in the code. This
should clear up an ambiguity between bank snapshots and snapshot
archives.
Michael
Jeff Washington (jwash)
carllin
Brooks Prumo
Michael Vines
sakridge
behzad nouri
Sean Young
Tyera Eulberg
Brooks Prumo
Jeff Biseda
Lijun Wang
Justin Starry
Trent Nelson
Jack May
Tao Zhu
Ryo Onodera