In order to avoid full replays when validating attestations hailing from
untaken forks, it's better to keep shufflings separate from `EpochRef`
and perform a lookahead on the shuffling when processing the block that
determines them.
This also helps performance in the case where REST clients are trying to
perform lookahead on attestation duties and decreases memory usage by
sharing shufflings between EpochRef instances of the same dependent
root.
The justified and finalized `Checkpoint` are frequently passed around
together. This introduces a new `FinalityCheckpoint` data structure that
combines them into one.
Due to the large usage of this structure in fork choice, also took this
opportunity to update fork choice tests to the latest v1.2.0-rc.1 spec.
Many additional tests enabled, some need more work, e.g. EL mock blocks.
Also implemented `discard_equivocations` which was skipped in #3661,
and improved code reuse across fork choice logic while at it.
* optimistic sync
* flag that initially loaded blocks from database might need execution block root filled in
* return optimistic status in REST calls
* refactor blockslot pruning
* ensure beacon_blocks_by_{root,range} do not provide optimistic blocks
* handle forkchoice head being pre-merge with block being postmerge
* re-enable blocking head updates on validator duties
* fix is_optimistic_candidate_block per spec; don't crash with nil future
* fix is_optimistic_candidate_block per spec; don't crash with nil future
* mark blocks sans execution payloads valid during head update
Some upstream repos still need fixes, but this gets us close enough that
style hints can be enabled by default.
In general, "canonical" spellings are preferred even if they violate
nep-1 - this applies in particular to spec-related stuff like
`genesis_validators_root` which appears throughout the codebase.
* harden validator API against pre-finalized slot requests
* check `syncHorizon` when responding to validator api requests too far
from `head`
* limit state-id based requests to one epoch ahead of `head`
* put historic data bounds on block/attestation/etc validator production API, preventing them from being used with already-finalized slots
* add validator block smoke tests
* make rest test create a new genesis with the tests running roughly in
the first epoch to allow testing a few more boundary conditions
* era: load blocks and states
Era files contain finalized history and can be thought of as an
alternative source for block and state data that allows clients to avoid
syncing this information from the P2P network - the P2P network is then
used to "top up" the client with the most recent data. They can be
freely shared in the community via whatever means (http, torrent, etc)
and serve as a permanent cold store of consensus data (and, after the
merge, execution data) for history buffs and bean counters alike.
This PR gently introduces support for loading blocks and states in two
cases: block requests from rest/p2p and frontfilling when doing
checkpoint sync.
The era files are used as a secondary source if the information is not
found in the database - compared to the database, there are a few key
differences:
* the database stores the block indexed by block root while the era file
indexes by slot - the former is used only in rest, while the latter is
used both by p2p and rest.
* when loading blocks from era files, the root is no longer trivially
available - if it is needed, it must either be computed (slow) or cached
(messy) - the good news is that for p2p requests, it is not needed
* in era files, "framed" snappy encoding is used while in the database
we store unframed snappy - for p2p2 requests, the latter requires
recompression while the former could avoid it
* front-filling is the process of using era files to replace backfilling
- in theory this front-filling could happen from any block and
front-fills with gaps could also be entertained, but our backfilling
algorithm cannot take advantage of this because there's no (simple) way
to tell it to "skip" a range.
* front-filling, as implemented, is a bit slow (10s to load mainnet): we
load the full BeaconState for every era to grab the roots of the blocks
- it would be better to partially load the state - as such, it would
also be good to be able to partially decompress snappy blobs
* lookups from REST via root are served by first looking up a block
summary in the database, then using the slot to load the block data from
the era file - however, there needs to be an option to create the
summary table from era files to fully support historical queries
To test this, `ncli_db` has an era file exporter: the files it creates
should be placed in an `era` folder next to `db` in the data directory.
What's interesting in particular about this setup is that `db` remains
as the source of truth for security purposes - it stores the latest
synced head root which in turn determines where a node "starts" its
consensus participation - the era directory however can be freely shared
between nodes / people without any (significant) security implications,
assuming the era files are consistent / not broken.
There's lots of future improvements to be had:
* we can drop the in-memory `BlockRef` index almost entirely - at this
point, resident memory usage of Nimbus should drop to a cool 500-600 mb
* we could serve era files via REST trivially: this would drop backfill
times to whatever time it takes to download the files - unlike the
current implementation that downloads block by block, downloading an era
at a time almost entirely cuts out request overhead
* we can "reasonably" recreate detailed state history from almost any
point in time, turning an O(slot) process into O(1) effectively - we'll
still need caches and indices to do this with sufficient efficiency for
the rest api, but at least it cuts the whole process down to minutes
instead of hours, for arbitrary points in time
* CI: ignore failures with Nim-1.6 (temporary)
* test fixes
Co-authored-by: Ștefan Talpalaru <stefantalpalaru@yahoo.com>
Gracefully handles the new failure modes recently introduced to the DAG
as part of https://github.com/status-im/nimbus-eth2/pull/3513
Data that is deemed to exist but fails to load leads to an error log to
avoid suppressing logic errors accidentally. In `verifyFinalization`
mode, the assertions remain active.
Up til now, the block dag has been using `BlockRef`, a structure adapted
for a full DAG, to represent all of chain history. This is a correct and
simple design, but does not exploit the linearity of the chain once
parts of it finalize.
By pruning the in-memory `BlockRef` structure at finalization, we save,
at the time of writing, a cool ~250mb (or 25%:ish) chunk of memory
landing us at a steady state of ~750mb normal memory usage for a
validating node.
Above all though, we prevent memory usage from growing proportionally
with the length of the chain, something that would not be sustainable
over time - instead, the steady state memory usage is roughly
determined by the validator set size which grows much more slowly. With
these changes, the core should remain sustainable memory-wise post-merge
all the way to withdrawals (when the validator set is expected to grow).
In-memory indices are still used for the "hot" unfinalized portion of
the chain - this ensure that consensus performance remains unchanged.
What changes is that for historical access, we use a db-based linear
slot index which is cache-and-disk-friendly, keeping the cost for
accessing historical data at a similar level as before, achieving the
savings at no percievable cost to functionality or performance.
A nice collateral benefit is the almost-instant startup since we no
longer load any large indicies at dag init.
The cost of this functionality instead can be found in the complexity of
having to deal with two ways of traversing the chain - by `BlockRef` and
by slot.
* use `BlockId` instead of `BlockRef` where finalized / historical data
may be required
* simplify clearance pre-advancement
* remove dag.finalizedBlocks (~50:ish mb)
* remove `getBlockAtSlot` - use `getBlockIdAtSlot` instead
* `parent` and `atSlot` for `BlockId` now require a `ChainDAGRef`
instance, unlike `BlockRef` traversal
* prune `BlockRef` parents on finality (~200:ish mb)
* speed up ChainDAG init by not loading finalized history index
* mess up light client server error handling - this need revisiting :)
One more step on the journey to reduce `BlockRef` usage across the
codebase - this one gets rid of `StateData` whose job was to keep track
of which block was last assigned to a state - these duties have now been
taken over by `latest_block_root`, a fairly recent addition that
computes this block root from state data (at a small cost that should be
insignificant)
99% mechanical change.
* fewer deps on `BlockRef` traversal in anticipation of pruning
* allows identifying EpochRef:s by their shuffling as a first step of
* tighten error handling around missing blocks
using the zero hash for signalling "missing block" is fragile and easy
to miss - with checkpoint sync now, and pruning in the future, missing
blocks become "normal".
Light clients require full nodes to serve additional data so that they
can stay in sync with the network. This patch adds a new launch option
`--import-light-client-data` to configure what data to make available.
For now, data is only kept in memory; it is not persisted at this time.
Note that data is only locally collected, a separate patch is needed to
actually make it availble over the network. `--serve-light-client-data`
will be used for serving data, but is not functional yet outside tests.
Can't apply a phase0 block to a later phase state and vice versa.
Since instantiation has been a topic, pre/post c file size:
```
424K @mspec@sstate_transition.nim.c
892K @mspec@sstate_transition_block.nim.c
```
```
288K @mspec@sstate_transition.nim.c
880K @mspec@sstate_transition_block.nim.c
```
This PR names and documents the concept of the archive: a range of slots
for which we have degraded functionality in terms of historical access -
in particular:
* we don't support rewinding to states in this range
* we don't keep an in-memory representation of the block dag
The archive de-facto exists in a trusted-node-synced node, but this PR
gives it a name and drops the in-memory digest index.
In order to satisfy `GetBlocksByRange` requests, we ensure that we have
blocks for the entire archive period via backfill. Future versions may
relax this further, adding a "pre-archive" period that is fully pruned.
During by-slot searches in the archive (both for libp2p and rest
requests), an extra database lookup is used to covert the given `slot`
to a `root` - future versions will avoid this using era files which
natively are indexed by `slot`. That said, the lookup is quite
fast compared to the actual block loading given how trivial the table
is - it's hard to measure, even.
A collateral benefit of this PR is that checkpoint-synced nodes will see
100-200MB memory usage savings, thanks to the dropped in-memory cache -
future pruning work will bring this benefit to full nodes as well.
* document chaindag storage architecture and assumptions
* look up parent using block id instead of full block in clearance
(future-proofing the code against a future in which blocks come from era
files)
* simplify finalized block init, always writing the backfill portion to
db at startup (to ensure lookups work as expected)
* preallocate some extra memory for finalized blocks, to avoid immediate
realloc
* Store finalized block roots in database (3s startup)
When the chain has finalized a checkpoint, the history from that point
onwards becomes linear - this is exploited in `.era` files to allow
constant-time by-slot lookups.
In the database, we can do the same by storing finalized block roots in
a simple sparse table indexed by slot, bringing the two representations
closer to each other in terms of conceptual layout and performance.
Doing so has a number of interesting effects:
* mainnet startup time is improved 3-5x (3s on my laptop)
* the _first_ startup might take slightly longer as the new index is
being built - ~10s on the same laptop
* we no longer rely on the beacon block summaries to load the full dag -
this is a lot faster because we no longer have to look up each block by
parent root
* a collateral benefit is that we no longer need to load the full
summaries table into memory - we get the RSS benefits of #3164 without
the CPU hit.
Other random stuff:
* simplify forky block generics
* fix withManyWrites multiple evaluation
* fix validator key cache not being updated properly in chaindag
read-only mode
* drop pre-altair summaries from `kvstore`
* recreate missing summaries from altair+ blocks as well (in case
database has lost some to an involuntary restart)
* print database startup timings in chaindag load log
* avoid allocating superfluos state at startup
* use a recursive sql query to load the summaries of the unfinalized
blocks
* limit by-root requests to non-finalized blocks
Presently, we keep a mapping from block root to `BlockRef` in memory -
this has simplified reasoning about the dag, but is not sustainable with
the chain growing.
We can distinguish between two cases where by-root access is useful:
* unfinalized blocks - this is where the beacon chain is operating
generally, by validating incoming data as interesting for future fork
choice decisions - bounded by the length of the unfinalized period
* finalized blocks - historical access in the REST API etc - no bounds,
really
In this PR, we limit the by-root block index to the first use case:
finalized chain data can more efficiently be addressed by slot number.
Future work includes:
* limiting the `BlockRef` horizon in general - each instance is 40
bytes+overhead which adds up - this needs further refactoring to deal
with the tail vs state problem
* persisting the finalized slot-to-hash index - this one also keeps
growing unbounded (albeit slowly)
Anyway, this PR easily shaves ~128mb of memory usage at the time of
writing.
* No longer honor `BeaconBlocksByRoot` requests outside of the
non-finalized period - previously, Nimbus would generously return any
block through this libp2p request - per the spec, finalized blocks
should be fetched via `BeaconBlocksByRange` instead.
* return `Opt[BlockRef]` instead of `nil` when blocks can't be found -
this becomes a lot more common now and thus deserves more attention
* `dag.blocks` -> `dag.forkBlocks` - this index only carries unfinalized
blocks from now - `finalizedBlocks` covers the other `BlockRef`
instances
* in backfill, verify that the last backfilled block leads back to
genesis, or panic
* add backfill timings to log
* fix missing check that `BlockRef` block can be fetched with
`getForkedBlock` reliably
* shortcut doppelganger check when feature is not enabled
* in REST/JSON-RPC, fetch blocks without involving `BlockRef`
* fix dag.blocks ref
Time in the beacon chain is expressed relative to the genesis time -
this PR creates a `beacon_time` module that collects helpers and
utilities for dealing the time units - the new module does not deal with
actual wall time (that's remains in `beacon_clock`).
Collecting the time related stuff in one place makes it easier to find,
avoids some circular imports and allows more easily identifying the code
actually needs wall time to operate.
* move genesis-time-related functionality into `spec/beacon_time`
* avoid using `chronos.Duration` for time differences - it does not
support negative values (such as when something happens earlier than it
should)
* saturate conversions between `FAR_FUTURE_XXX`, so as to avoid
overflows
* fix delay reporting in validator client so it uses the expected
deadline of the slot, not "closest wall slot"
* simplify looping over the slots of an epoch
* `compute_start_slot_at_epoch` -> `start_slot`
* `compute_epoch_at_slot` -> `epoch`
A follow-up PR will (likely) introduce saturating arithmetic for the
time units - this is merely code moves, renames and fixing of small
bugs.
With checkpoint sync in particular, and state pruning in the future,
loading states or state-dependent data may fail. This PR adjusts the
code to allow this to be handled gracefully.
In particular, the new availability assumption is that states are always
available for the finalized checkpoint and newer, but may fail for
anything older.
The `tail` remains the point where state loading de-facto fails, meaning
that between the tail and the finalized checkpoint, we can still get
historical data (but code should be prepared to handle this as an
error).
However, to harden the code against long replays, several operations
which are assumed to work only with non-final data (such as gossip
verification and validator duties) now limit their search horizon to
post-finalized data.
* harden several state-dependent operations by logging an error instead
of introducing a panic when state loading fails
* `withState` -> `withUpdatedState` to differentiate from the other
`withState`
* `updateStateData` can now fail if no state is found in database - it
is also hardened against excessively long replays
* `getEpochRef` can now fail when replay fails
* reject blocks with invalid target root - they would be ignored
previously
* fix recursion bug in `isProposed`
With the right sequence of events (for example a REST request or a
validation), it can happen that the first traversal across a state
checkpoint boundary is done without storing that state on disk - this
causes problens when replaying states, because now states may be missing
from the database.
Here, we simply avoid using the caches when advancing a state that will
go into the database, ensuring that the information lost during caching
always is permanently stored.
* fix recursion bug in `isProposed`
Introduced in #3171, it turns out we can just follow the block headers
to achieve the same effect
* leaves the constant in the code so as to avoid confusion when reading
database that had the constant written (such as the fleet nodes and
other unstable users)
In the ChainDAG, 3 block pointers are kept: genesis, tail and head. This
PR adds one more block pointer: the backfill block which represents the
block that has been backfilled so far.
When doing a checkpoint sync, a random block is given as starting point
- this is the tail block, and we require that the tail block has a
corresponding state.
When backfilling, we end up with blocks without corresponding states,
hence we cannot use `tail` as a backfill pointer - there is no state.
Nonetheless, we need to keep track of where we are in the backfill
process between restarts, such that we can answer GetBeaconBlocksByRange
requests.
This PR adds the basic support for backfill handling - it needs to be
integrated with backfill sync, and the REST API needs to be adjusted to
take advantage of the new backfilled blocks when responding to certain
requests.
Future work will also enable moving the tail in either direction:
* pruning means moving the tail forward in time and removing states
* backwards means recreating past states from genesis, such that
intermediate states are recreated step by step all the way to the tail -
at that point, tail, genesis and backfill will match up.
* backfilling is done when backfill != genesis - later, this will be the
WSS checkpoint instead
* move quarantine outside of chaindag
The quarantine has been part of the ChainDAG for the longest time, but
this design has a few issues:
* the function in which blocks are verified and added to the dag becomes
reentrant and therefore difficult to reason about - we're currently
using a stateful flag to work around it
* quarantined blocks bypass the processing queue leading to a processing
stampede
* the quarantine flow is unsuitable for orphaned attestations - these
should also should be quarantined eventually
Instead of processing the quarantine inside ChainDAG, this PR moves
re-queueing to `block_processor` which already is responsible for
dealing with follow-up work when a block is added to the dag
This sets the stage for keeping attestations in the quarantine as well.
Also:
* make `BlockError` `{.pure.}`
* avoid use of `ValidationResult` in block clearance (that's for gossip)
* fix stack overflow crash in REST/debug/getStateV2
* introduce `ForkyXxx` for generic type matching of `Xxx` across
branches (SomeHashedBeaconState -> ForkyHashedBeaconState et al) -
`Some` is already used for other types of type classes
* consolidate function naming in BeaconChainDB, use some generics
* import `forks.nim` from other spec modules and move `Forked*` helpers
around to resolve circular imports
* remove `ForkedBeaconState`, use `ForkedHashedBeaconState` throughout
(less data shuffling between the types)
* fix several cases of states being stored on stack in tests, causing
random failures on some platforms
* remove reading json support from ncli - this should be ported to the
rest json reading instead (doesn't currently work because stack sizes)
* Placing callbacks into strategic places.
* Initial events call implementation.
* Post rebase fixes.
* Change addSyncContribution() implementation.
* Add `attestation-sent` event.
Remove gcsafe, raises from callbacks implementations.
Move `attestation-received` fire at the end of attestation processing.
* Address review comments.
* Add parallel attestation verification
* Update tests, batchVerify doesn't use the threadpool with only single core (nim-blscurve update)
* bump nim-blscurve
* Debug info for failing eth2 test vectors
* remove submodule eth2-testnets
* verbose debugging of make failure on Windows (libbacktrace?)
* Remove CI debug mode
* initialization convention
* Fix new altair tests
* reorganize ssz dependencies
This PR continues the work in
https://github.com/status-im/nimbus-eth2/pull/2646,
https://github.com/status-im/nimbus-eth2/pull/2779 as well as past
issues with serialization and type, to disentangle SSZ from eth2 and at
the same time simplify imports and exports with a structured approach.
The principal idea here is that when a library wants to introduce SSZ
support, they do so via 3 files:
* `ssz_codecs` which imports and reexports `codecs` - this covers the
basic byte conversions and ensures no overloads get lost
* `xxx_merkleization` imports and exports `merkleization` to specialize
and get access to `hash_tree_root` and friends
* `xxx_ssz_serialization` imports and exports `ssz_serialization` to
specialize ssz for a specific library
Those that need to interact with SSZ always import the `xxx_` versions
of the modules and never `ssz` itself so as to keep imports simple and
safe.
This is similar to how the REST / JSON-RPC serializers are structured in
that someone wanting to serialize spec types to REST-JSON will import
`eth2_rest_serialization` and nothing else.
* split up ssz into a core library that is independendent of eth2 types
* rename `bytes_reader` to `codec` to highlight that it contains coding
and decoding of bytes and native ssz types
* remove tricky List init overload that causes compile issues
* get rid of top-level ssz import
* reenable merkleization tests
* move some "standard" json serializers to spec
* remove `ValidatorIndex` serialization for now
* remove test_ssz_merkleization
* add tests for over/underlong byte sequences
* fix broken seq[byte] test - seq[byte] is not an SSZ type
There are a few things this PR doesn't solve:
* like #2646 this PR is weak on how to handle root and other
dontSerialize fields that "sometimes" should be computed - the same
problem appears in REST / JSON-RPC etc
* Fix a build problem on macOS
* Another way to fix the macOS builds
Co-authored-by: Zahary Karadjov <zahary@gmail.com>
The spec imports are a mess to work with, so this branch cleans them up
a bit to ensure that we avoid generic sandwitches and that importing
stuff generally becomes easier.
* reexport crypto/digest/presets because these are part of the public
symbol set of the rest of the spec types
* don't export `merge` types from `base` - this causes circular deps
* fix circular deps in `ssz/spec_types` - this is the first step in
disentangling ssz from spec
* be explicit about phase0 vs altair - longer term, `altair` will become
the "natural" type set, then merge and so on, so no point in giving
`phase0` special preferential treatment
Simpler module name for stuff that covers forks
* check that runtime config matches database state
* also include some assorted altair cleanups
* use "standard" genesis fork in local testnet to work around missing
runtime config support
* update validator key cache on startup
Versions prior to 1.1.0 do not write a validator key cache at all.
Versions from 1.4.0 and upwards require an immutable validator key cache
to verify blocks - normally, block verification fills the cache but that
assumes that at least one block was verified by a version that has the
key cache.
Taken together, this breaks direct upgrades from anything <1.1.0 to
1.4.0.
The fix is simply to refresh fill the cache from an existing state on
startup.
* also log serious block validation failures at info level
* Implement split preset/config support
This is the initial bulk refactor to introduce runtime config values in
a number of places, somewhat replacing the existing mechanism of loading
network metadata.
It still needs more work, this is the initial refactor that introduces
runtime configuration in some of the places that need it.
The PR changes the way presets and constants work, to match the spec. In
particular, a "preset" now refers to the compile-time configuration
while a "cfg" or "RuntimeConfig" is the dynamic part.
A single binary can support either mainnet or minimal, but not both.
Support for other presets has been removed completely (can be readded,
in case there's need).
There's a number of outstanding tasks:
* `SECONDS_PER_SLOT` still needs fixing
* loading custom runtime configs needs redoing
* checking constants against YAML file
* yeerongpilly support
`build/nimbus_beacon_node --network=yeerongpilly --discv5:no --log-level=DEBUG`
* load fork epoch from config
* fix fork digest sent in status
* nicer error string for request failures
* fix tools
* one more
* fixup
* fixup
* fixup
* use "standard" network definition folder in local testnet
Files are loaded from their standard locations, including genesis etc,
to conform to the format used in the `eth2-networks` repo.
* fix launch scripts, allow unknown config values
* fix base config of rest test
* cleanups
* bundle mainnet config using common loader
* fix spec links and names
* only include supported preset in binary
* drop yeerongpilly, add altair-devnet-0, support boot_enr.yaml
* Implement the new Altair req/resp protocols
Also fixes the altair message-id computation by providing the correct
forkdigest prefix in `isAltairTopic`.
Co-authored-by: Tanguy Cizain <tanguycizain@gmail.com>
* add Altair support to the block quarantine
* switch some spec/datatypes imports to spec/datatypes/base
* add Altair support to block_clearance
* allow runtime configuration of Altair transition slot
* enable Altair in block_sim, including in CI
* remove false OnBlockAdded dependency on phase.HashedBeaconState
* introduce altair data types into block_clearance; update some alpha.6 spec refs to alpha.7; add get_active_validator_indices_len ForkedHashedBeaconState wrapper
* switch many modules from using datatypes (with phase0 states/blocks) to datatypes/base (fork-independent); update spec refs from alpha.6 to alpha.7 and remove rm'd G2_POINT_AT_INFINITY
* switch more modules from using datatypes (with phase0 states/blocks) to datatypes/base (fork-independent); update spec refs from alpha.6 to alpha.7
* remove unnecessary phase0-only wrapper of get_attesting_indices(); allow signatures_batch to process either fork; remove O(n^2) nested loop in process_inactivity_updates(); add altair support to getAttestationsforTestBlock()
* add Altair versions of asSigVerified(), asTrusted(), and makeBeaconBlock()
* fix spec URL to be Altair for Altair makeBeaconBlock()
* use ForkedHashedBeaconState in StateData
* fix FAR_FUTURE_EPOCH -> slot overflow; almost always use assign()
* avoid stack allocation in maybeUpgradeStateToAltair()
* create and use dispatch functions for check_attester_slashing(), check_proposer_slashing(), and check_voluntary_exit()
* use getStateRoot() instead of various state.data.hbsPhase0.root
* remove withStateVars.hashedState(), which doesn't work as a design anymore
* introduce spec/datatypes/altair into beacon_chain_db
* fix inefficient codegen for getStateField(largeStateField)
* state_transition_slots() doesn't either need/use blocks or runtime presets
* combine process_slots(HBS)/state_transition_slots(HBS) which differ only in last-slot htr optimization
* getStateField(StateData, ...) was replaced by getStateField(ForkedHashedBeaconState, ...)
* fix rollback
* switch some state_transition(), process_slots, makeTestBlocks(), etc to use ForkedHashedBeaconState
* remove state_transition(phase0.HashedBeaconState)
* remove process_slots(phase0.HashedBeaconState)
* remove state_transition_block(phase0.HashedBeaconState)
* remove unused callWithBS(); separate case expression from if statement
* switch back from nested-ref-object construction to (ref Foo)(Bar())
* write uncompressed validator keys to database
Loading 150k+ validator keys on startup in compressed format takes a lot
of time - better store them in uncompressed format which makes behaviour
just after startup faster / more predictable.
* refactor cached validator key access
* fix isomorphic cast to work with non-var instances
* remove cooked pubkey cache - directly use database cache in chaindag
as well (one less cache to keep in sync)
* bump blscurve, introduce loadValid for known-to-be-valid keys
tersec
Jacek Sieka
Miran
Etan Kissling
Eugene Kabanov
Mamy Ratsimbazafy
zah