Commit Graph

109 Commits

Author SHA1 Message Date
Brooks Prumo 4e3818e5c1
Add CrdsData::IncrementalSnapshotHashes (#20374) 2021-10-05 09:57:46 -05:00
Brooks Prumo 5d141fe01d
Rename CRDS SnapshotHash to SnapshotHashes (#20421) 2021-10-04 19:03:28 -05:00
carllin ee8621a8bd
Add metric measuring number of successfully inserted push messages (#20275)
* Add number of successfully inserted push messages
2021-09-28 21:41:17 -07:00
behzad nouri 43ed727ba7
reverts #17542 (#20259)
https://github.com/solana-labs/solana/pull/17542
excludes caller's crds values from pull responses.

Reverting that commit so that when a (staked) node restarts, it can
obtain its crds values before restart from other nodes.
2021-09-27 22:03:26 +00:00
sakridge 013e1d9d49
Limit transaction forwarding from banking_stage (#19940) 2021-09-21 08:49:41 -07:00
sakridge 44c8b1bca2
Remove clippy (#19793) 2021-09-13 20:08:28 -07:00
behzad nouri d7051b0d21
adds logs when push-vote panics with invalid vote-index (#19485)
In order to debug this panic on the clusters:

  panicked at 'assertion failed: (vote_index as usize) <
  MAX_LOCKOUT_HISTORY', core/src/cluster_info.rs:1012:9
2021-08-31 12:15:07 +00:00
behzad nouri 6909a79b6f
removes require-stake-for-gossip feature (#19476)
The feature is already activated on all clusters.
2021-08-27 21:17:15 +00:00
behzad nouri 3efccbffab sends shreds (instead of packets) to retransmit stage
Working towards channelling through shreds recovered from erasure codes
to retransmit stage.
2021-08-17 13:44:10 +00:00
behzad nouri 140abec6ef
exempts node-instances from shred-version check (#19190)
Clusters are kept separate using the shred-versions obtained from
contact-infos. However, this mechanism breaks if there are 2 instances
of the same identity key running on different clusters, because then one
of the two contact-infos have the right shred-version.
If a node has the contact-info with the matching shred-version, then it
will pass all associated crds values even if they belong to the other
instance. So the shred-version check breaks.
As a result we cannot support 2 instances of the same identity key
running on different clusters. To prevent that, this commit is exempting
node-instances from shred-version check so that they are always
propagated across clusters and halt one of the running duplicate
instances.
2021-08-14 00:47:44 +00:00
behzad nouri 7a789e0763
filters for recent contact-infos when checking for live stake (#19204)
Contact-infos are saved to disk:
https://github.com/solana-labs/solana/blob/9dfeee299/gossip/src/cluster_info.rs#L1678-L1683

and restored on validator start-up:
https://github.com/solana-labs/solana/blob/9dfeee299/core/src/validator.rs#L450

Staked nodes entries will not expire until an epoch after. So when the
validator checks for online stake it is erroneously picking up
contact-infos restored from disk, which breaks the entire
wait-for-supermajority logic:
https://github.com/solana-labs/solana/blob/9dfeee299/core/src/validator.rs#L1515-L1561

This commit adds an extra check for the age of contact-info entries and
filters out old ones.
2021-08-13 12:12:40 +00:00
behzad nouri f302774cf7
implements copy-on-write for staked-nodes (#19090)
Bank::staked_nodes and Bank::epoch_staked_nodes redundantly clone
staked-nodes HashMap even though an immutable reference will suffice:
https://github.com/solana-labs/solana/blob/a9014cece/runtime/src/vote_account.rs#L77

This commit implements copy-on-write semantics for staked-nodes by
wrapping the underlying HashMap in Arc<...>.
2021-08-10 12:59:12 +00:00
Justin Starry 8817f59b6e
Version transaction message and add new message format (#18725)
* Version transaction message and add new message format

* Update abi digest due to message path change

* Update v0.rs

Fix comment

* Update original.rs

* Update message versions name and address map indexes field name

* s/original/legacy

* update comment

* cargo fmt

* Update abi digest due to legacy rename
2021-08-09 22:03:39 -07:00
behzad nouri 049fb0417f
allows sendmmsg api taking owned values (as well as references) (#18999)
Current signature of api in sendmmsg requires a slice of inner
references:
https://github.com/solana-labs/solana/blob/fe1ee4980/streamer/src/sendmmsg.rs#L130-L152

That forces the call-site to convert owned values to references even
though doing so is redundant and adds an extra level of indirection:
https://github.com/solana-labs/solana/blob/fe1ee4980/core/src/repair_service.rs#L291

This commit expands the api using AsRef and Borrow traits to allow
calling the method with owned values (as well as references like
before).
2021-07-30 20:58:49 +00:00
behzad nouri 81026f9ea5
passes through --allow-private-addr to validators in system perf tests (#18876) 2021-07-29 19:04:45 +00:00
behzad nouri f1198fc6d5
filters crds values in parallel when responding to gossip pull-requests (#18877)
When responding to gossip pull-requests, filter_crds_values takes a lot of time
while holding onto read-lock:
https://github.com/solana-labs/solana/blob/f51d64868/gossip/src/crds_gossip_pull.rs#L509-L566

This commit will filter-crds-values in parallel using rayon thread-pools.
2021-07-26 17:13:11 +00:00
behzad nouri d2d5f36a3c
adds validator flag to allow private ip addresses (#18850) 2021-07-23 15:25:03 +00:00
carllin 588c0464b8
Add sampling logic and DuplicateSlotRepairStatus module (#18721) 2021-07-21 11:15:08 -07:00
behzad nouri bbd22f06f4
implements generic lookups into gossip crds table (#18765)
This commit adds CrdsEntry trait which allows generic lookups into crds
table. For example to get ContactInfo or LowestSlot associated with a
Pubkey, the lookup code would be respectively:
   crds.get::<&ContactInfo>(pubkey)
   crds.get::<&LowestSlot>(pubkey)
2021-07-21 12:16:26 +00:00
Justin Starry 207c90bd8b
Shorten long SerializeWith type paths in abi digest (#18734) 2021-07-20 08:59:50 -05:00
behzad nouri 8da261cf5c
locks crds only once in ClusterInfo::repair_peers (#18752)
ClusterInfo::repair_peers locks crds table twice, and shows performance
regression if the RwLock is not reader-preferred:
https://github.com/solana-labs/solana/blob/269028360/gossip/src/cluster_info.rs#L1188-L1210
2021-07-18 16:55:58 +00:00
behzad nouri e316586516 excludes private ip addresses 2021-07-16 20:05:48 -06:00
Jeff Biseda ae5ad5cf9b
sendmmsg cleanup #18589
Rationalize usage of sendmmsg(2). Skip packets which failed to send and track failures.
2021-07-16 14:36:49 -07:00
Brian Anderson 37ee0b5599
Eliminate doc warnings and fix some markdown (#18566)
* Fix link target in doc comment

* Fix formatting of log examples in process_instruction

* Fix doc markdown in solana-gossip

* Fix doc markdown in solana-runtime

* Escape square braces in doc comments to avoid warnings

* Surround 'account references' doc items in code spans to avoid warnings

* Fix code block in loader_upgradeable_instruction

* Fix doctest for loader_upgradable_instruction
2021-07-16 00:40:07 +00:00
behzad nouri cf31afdd6a
makes CrdsGossip thread-safe (#18615) 2021-07-14 22:27:17 +00:00
sakridge 7f2254225e
Move entry/poh to own crate to speed up poh bench build (#18225) 2021-07-14 14:16:29 +02:00
behzad nouri c90af3cd63
removes id from push_lowest_slot args (#18645)
push_lowest_slot cannot sign the new crds-value unless the id (pubkey)
argument passed-in is the same pubkey as in ClusterInfo::keypair(), in
which case the id argument is redundant:
https://github.com/solana-labs/solana/blob/bb41cf346/gossip/src/cluster_info.rs#L824-L845

Additionally, the lookup is done with self.id(), but insert is done with
the id argument, which is logically a bug.
2021-07-13 22:32:59 +00:00
behzad nouri 90f8cf0920
makes CrdsGossipPush thread-safe (#18581) 2021-07-13 14:04:25 +00:00
behzad nouri e7a1f2c9b0
makes CrdsGossipPull thread-safe (#18578) 2021-07-11 15:32:10 +00:00
carllin 175083c4c1
Add updated duplicate broadcast test (#18506) 2021-07-10 22:22:07 -07:00
behzad nouri 918b5c28b2
removes redundant (mutable) self receivers (#18574) 2021-07-10 22:16:33 +00:00
behzad nouri fd9c10c2e2
adds a generic implementation of Gossip{Read,Write}Lock (#18559) 2021-07-10 14:13:52 +00:00
behzad nouri 4e1333fbe6
removes id and shred_version from CrdsGossip (#18505)
ClusterInfo is the gateway to CrdsGossip function calls, and it already
has node's pubkey and shred version (full ContactInfo and Keypair in
fact).
Duplicating these data in CrdsGossip adds redundancy and possibility for
bugs should they not be consistent with ClusterInfo.
2021-07-09 13:10:08 +00:00
behzad nouri 27cc7577a1
skips process_push_message for local messages (#18493)
received_cache is not relevant for local messages, and does not need to
be updated:
https://github.com/solana-labs/solana/blob/92c5cdab6/gossip/src/crds_gossip_push.rs#L166-L189
2021-07-09 01:42:13 +00:00
Michael Vines 1e0942e900 Rename ClusterInfo::send_vote to ClusterInfo::send_transaction 2021-07-07 15:51:14 -07:00
Justin Starry 92c5cdab62
Fix cargo check (#18499) 2021-07-07 14:21:08 -05:00
behzad nouri dba42c57b4
implements an unbiased weighted shuffle using binary indexed tree (#18343)
Current implementation of weighted_shuffle:
https://github.com/solana-labs/solana/blob/b08f8bd1b/gossip/src/weighted_shuffle.rs#L11-L37
uses a heuristic which results in biased samples.

For example, if the weights are [1, 10, 100], then the 3rd index should
come first 100 times more often than the 1st index. However,
weighted_shuffle is picking the 3rd index 200+ times more often than the
1st index, showing a disproportional bias in favor of higher weights.

This commit implements weighted shuffle using binary indexed tree to
maintain cumulative sum of weights while sampling. The resulting samples
are demonstrably unbiased and precisely proportional to the weights.

Additionally the iterator interface allows to skip computations when
not all indices are processed.

Of the use cases of weighted_shuffle, changing turbine code requires
feature-gating to keep the cluster in sync. That is not updated in
this commit, but can be done together with future updates to turbine.
2021-07-07 14:14:43 +00:00
behzad nouri 04787be8b1
encapsulates turbine peers computations of broadcast & retransmit stages (#18238)
Broadcast stage and retransmit stage should arrange nodes on turbine
broadcast tree in exactly same order. Additionally any changes to this
ordering (e.g. updating how unstaked nodes are handled) requires feature
gating to keep the cluster in sync.

Current implementation is scattered out over several public methods and
exposes too much of implementation details (e.g. usize indices into
peers vector) which makes code changes and checking for feature
activations more difficult.

This commit encapsulates turbine peer computations into a new struct,
and only exposes two public methods, get_broadcast_peer and
get_retransmit_peers, for call-sites.
2021-07-07 00:35:25 +00:00
Michael Vines c17451ca73 Acquire instance read lock once 2021-07-01 17:50:04 -07:00
Michael Vines db3a9ae7fb Fully replace NodeInstance 2021-07-01 17:50:04 -07:00
Michael Vines 71efac46cb Hoist keypair() out of some loops 2021-07-01 17:50:04 -07:00
Michael Vines b6792a3328 Add ability to change the validator identity at runtime 2021-07-01 17:50:04 -07:00
Michael Vines bf157506e8 Remove id ref 2021-07-01 17:50:04 -07:00
Ashwin Sekar f4fb5de545
Consider all peers as potential candidates during pull-request in case of offline nodes (#18333)
* Try all peers during pull-request in case of offline nodes

* fix clippy err
2021-07-01 12:00:10 -07:00
behzad nouri 9d983a34a0
debug logs when crds table trim failed (#18307)
reports of this error being possibly spammy:
https://discord.com/channels/428295358100013066/689412830075551748/859441080054710293

The commit changes the log level to debug.
Additionally adding a new metric to understand the frequency of this error.
2021-06-29 19:39:46 +00:00
behzad nouri d7b8329b45
removes repeated calls to ClusterInfo::id in iterators and contact-info clone (#18174)
Calling ClusterInfo::id repeatedly in for loops or iterators is
inefficient, because it acquires a lock on ClusterInfo.my_contact_info,
and clones the entire contact-info.
2021-06-23 16:30:14 +00:00
behzad nouri 69a5f0e6cd
filters crds values obtained through gossip by their shred version (#18072)
filter_by_shred_version does not check the shred-version of the owner of
the crds-value. It only checks the shred-version of the node which is
relaying the value:
https://github.com/solana-labs/solana/blob/5cc073420/gossip/src/cluster_info.rs#L2274-L2289

So crds-values with different shred versions can still pass through this
function as long as they are relayed by a node with matching shred
version; and so, a single node can bridge different shred values
through-out the cluster.
2021-06-23 14:16:05 +00:00
Michael Vines 84b9de8c18 Shredder no longer holds a keypair 2021-06-21 21:29:52 -07:00
Michael Vines 553fc210f5 Remove duplicated id field 2021-06-21 21:29:52 -07:00
behzad nouri 598093b5db adds shred-version to ip-echo-server response
When starting a validator, the node initially joins gossip with
shred_verison = 0, until it adopts the entrypoint's shred-version:
https://github.com/solana-labs/solana/blob/9b182f408/validator/src/main.rs#L417

Depending on the load on the entrypoint, this adopting entrypoint
shred-version through gossip sometimes becomes very slow, and causes
several problems in gossip because we have to partially support
shred_version == 0 which is a source of leaking crds values from one
cluster to another. e.g. see
https://github.com/solana-labs/solana/pull/17899
and the other linked issues there.

In order to remove shred_version == 0 from gossip, this commit adds
shred-version to ip-echo-server response. Once the entrypoints are
updated, on validator start-up, if --expected_shred_version is not
specified we will obtain shred-version from the entrypoint using
ip-echo-server.
2021-06-21 19:37:16 +00:00