Log a "Trying..." message before each listener opens, to see if the
delay is inside Zebra, or in the test harness or OS.
Also report the configured and actual ports where possible, for better
diagnostics.
Design:
- Add a `PeerAddrState` to each `MetaAddr`
- Use a single peer set for all peers, regardless of state
- Implement time-based liveness as an `AddressBook` method, rather than
a `PeerAddrState` variant
- Delete `AddressBook.by_state`
Implementation:
- Simplify `AddressBook` changes using `update` and `take` modifier
methods
- Simplify the `AddressBook` iterator implementation, replacing it with
methods that are more obviously correct
- Consistently collect peer set metrics
Documentation:
- Expand and update the peer set documentation
We can optimise later, but for now we want simple code that is more
obviously correct.
* replace to_socket_addrs
* refactor `resolve()` into `resolve_host()`
* use `resolve_host()` to resolve config peers
* add DNS_LOOKUP_TIMEOUT constant
* don't block the main thread in initialize
* add hint for port error
* add issue filter for port panic
* add lock file hint
* add metrics endpoint port conflict hint
* add hint for tracing endpoint port conflict
* add acceptance test for resource conflics
* Split out common conflict test code into a function
* Add state, metrics, and tracing conflict tests
* Add a full set of stderr acceptance test functions
This change makes the stdout and stderr acceptance test interfaces
identical.
* move Zcash listener opening
* add todo about hint for disk full
* add constant for lock file
* match path in state cache
* don't match windows cache path
* Use Display for state path logs
Avoids weird escaping on Windows when using Debug
* Add Windows conflict error messages
* Turn PORT_IN_USE_ERROR into a regex
And add another alternative Windows-specific port error
Co-authored-by: teor <teor@riseup.net>
Co-authored-by: Jane Lusby <jane@zfnd.org>
## Motivation
This PR is motivated by the regression identified in https://github.com/ZcashFoundation/zebra/issues/1349. That PR notes that the metrics stopped working for most of the crates other than `zebrad`.
## Solution
This PR resolves the regression by deduplicating the `metrics` crate dependency. During a recent change we upgraded the metrics version in `zebrad` and a couple other of our crates, but we never updated the dependencies in `zebra-state`, `zebra-consensus`, or `zebra-network`. This caused the metrics macros to attempt to retrieve the current metrics exporter through the wrong function. We would install the metrics exporter in `0.13`, but then attempt to look it up through the `0.12` crate, which contains a different instance of the metrics exporter static variable which is unset. Doing this causes the metrics macros to return `None` for the current exporter after which they just silently give up.
## Related Issues
closes https://github.com/ZcashFoundation/zebra/issues/1349
## Follow Up Work
I noticed we have quite a few duplicate dependencies in our tree. We might be able to save some compilation time by auditing those and deduplicating them as much as possible.
- https://github.com/ZcashFoundation/zebra/issues/1582
Co-authored-by: teor <teor@riseup.net>
* warn: if there are no peers at all
* info: if there are no ready peers
* trace: the number of ready and unready peers for every request
Log at most one warn or info log per minute, to avoid flooding the
terminal with log lines. Suppress warn and info logs for the first
minute, while the peer set is starting up.
Closes#1183.
The peer set maintains a preselected ready service that it can use to
perform power-of-two-choices (p2c) routing of requests. Ready services
are stored by key (socket address) in an `IndexMap`, and the preselected
service is represented by an `Option<usize>` indexing that map. This
means that whenever the set of ready services changes (e.g., a service
is removed from the peer set, or a service is taken to be used to
process a request), the preselected index is invalidated. The original
P2C-only implementation maintained this invariant but did not document
it.
The change to inventory-based routing introduced a bug by failing to
maintain this invariant and appropriately invalidate the preselected
index. However, this was only noticeable approximately 1/N of the time
on the next request after an inventory-directed request, so the bug
occurred infrequently. Luckily, the use of `.expect` caused the bug to
be an immediate panic, making it possible to identify by inspecting all
uses of the ready service map.
This change is mostly mechanical, with the exception of the changes to the
`tower-batch` middleware. This middleware was adapted from `tower::buffer`,
and the `tower::buffer` code was changed to implement its own bounded queue,
because Tokio 0.3 removed the `mpsc::Sender::poll_send` method. See
ddc64e8d4d
for more context on the Tower changes. To match Tower as closely as possible
in order to be able to upstream `tower-batch`, those changes are copied from
`tower::Buffer` to `tower-batch`.
This addresses at least three pain points:
- we were affected by bugs that were already fixed in git, but not in
the released crate;
- we can use service combinators to transform requests and responses;
- we can use the hedge middleware.
The version in git is still marked as 0.3.1 but these changes will be
part of tower 0.4: https://github.com/tower-rs/tower/issues/431
The GetPeers requests sent while crawling the network are randomly
load-balanced over available peers. But at the very beginning, they may
be both routed to the same peer, causing network initialization to be
delayed while the second one times out (since zcashd only ever responds
to the first addr message).
Only sending one GetPeers request per candidate set update means we
crawl the network a little more slowly, but avoids hanging on start.
This commit makes several related changes to the network code:
- adds a `TransactionsByHash(HashSet<transaction::Hash>)` request and
`Transactions(Vec<Arc<Transaction>>)` response pair that allows
fetching transactions from a remote peer;
- adds a `PushTransaction(Arc<Transaction>)` request that pushes an
unsolicited transaction to a remote peer;
- adds an `AdvertiseTransactions(HashSet<transaction::Hash>)` request
that advertises transactions by hash to a remote peer;
- adds an `AdvertiseBlock(block::Hash)` request that advertises a block
by hash to a remote peer;
Then, it modifies the connection state machine so that outbound
requests to remote peers are handled properly:
- `TransactionsByHash` generates a `getdata` message and collects the
results, like the existing `BlocksByHash` request.
- `PushTransaction` generates a `tx` message, and returns `Nil` immediately.
- `AdvertiseTransactions` and `AdvertiseBlock` generate an `inv`
message, and return `Nil` immediately.
Next, it modifies the connection state machine so that messages
from remote peers generate requests to the inbound service:
- `getdata` messages generate `BlocksByHash` or `TransactionsByHash`
requests, depending on the content of the message;
- `tx` messages generate `PushTransaction` requests;
- `inv` messages generate `AdvertiseBlock` or `AdvertiseTransactions`
requests.
Finally, it refactors the request routing logic for the peer set to
handle advertisement messages, providing three routing methods:
- `route_p2c`, which uses p2c as normal (default);
- `route_inv`, which uses the inventory registry and falls back to p2c
(used for `BlocksByHash` or `TransactionsByHash`);
- `route_all`, which broadcasts a request to all ready peers (used for
`AdvertiseBlock` and `AdvertiseTransactions`).
* network: fix bug in inventory advertisement handling
The RFC https://zebra.zfnd.org/dev/rfcs/0003-inventory-tracking.html described
the use of a `broadcast` channel in place of an `mpsc` channel to get
ring-buffer behavior, keeping a bound on the size of the channel but dropping
old entries when the channel is full.
However, it didn't explicitly describe how this works (the `broadcast` channel
returns a `RecvError::Lagged(u64)` to inform receivers that they lost
messages), so the lag-handling wasn't implemented and I didn't notice in
review.
Instead, the ? operator bubbled the lag error all the way up from
`InventoryRegistry::poll_inventory` through `<PeerSet as Service>::poll_ready`
through various Tower wrappers to users of the peer set. The error propagation
is bad enough, because it caused client errors that shouldn't have happened,
but there's a worse interaction.
The `Service` contract distinguishes between request errors (from
`Service::call`, scoped to the request) and service errors (from
`Service::poll_ready`, scoped to the service). The `Service` contract
specifies that once a service returns an error from `poll_ready`, the service
can be assumed to be failed permanently.
I believe (but haven't tested or carefully worked through the details) that
this caused various tower middleware to report the entire peer set service as
permanently failed due to a transient inventory "error" (more of an indicator),
and I suspect that this is the cause of #1003, where all of the sync
component's requests end up failing because the peer set reported that it
failed permanently. I am able to reproduce #1003 locally before this change
and unable to reproduce it locally after this change, though I have not tested
exhaustively.
* network: add metric for dropped inventory advertisements
Co-authored-by: teor <teor@riseup.net>
Co-authored-by: teor <teor@riseup.net>
The relay flag in the version message is used in conjunction with BIP37 to
receive bloom-filtered transactions. When it is set to false, transactions are
not relayed until a bloom filter is set. Since we don't implement BIP37 (it's
not useful for shielded transactions), this means we'll never receive
transactions.
* network: move gossiped peer selection logic into address book.
* network: return BoxService from init.
* zebrad: add note on why we truncate thegossiped peer list
Co-authored-by: Jane Lusby <jlusby42@gmail.com>
* Remove unused .rustfmt.toml
Many of these options are never actually loaded by our CI because of a channel
mismatch, where they're not applied on stable but only on nightly (see the logs
from a rustfmt job). This means that we can get different settings when
running `cargo fmt` on the nightly and stable channels, which was causing a CI
failure on this PR. Reverting back to the default rustfmt settings avoids this
problem and keeps us in line with upstream rustfmt. There's no loss to us
since we were using the defaults anyways.
Co-authored-by: Jane Lusby <jlusby42@gmail.com>
Closes#536.
This removes:
- the user-agent (we can add a mechanism to specify extra BIP14 components later, if any users ask us for that feature);
- the EWMA parameters (these were put in the config just to avoid making a choice);
- the peer connection timeout (we can change the default value if anyone ever has a problem with it);
- the peer set request buffer size (setting this too low can make the application deadlock);
The new peer interval is left in.
Prior to this change, we required that services that are canceled do not
have a cancel handle in the `cancel_handles` list, based on the
assumption that the handle must have been removed in the process of
canceling this service.
This doesn't holding up though, because it is currently possible for us
to have the same peer connect to us multiple times, the second connect
removes the cancel handle of the original connect and inserts it's own
cancel handle in its place. In this scenario, when the first service is
polled for readiness it will see that it has been canceled and go to
clean itself up, but when it asserts that it doesn't have a cancel
handle it will see the cancel handle of the second connect event, which
uses the same key as the first connect, and fail its debug assertion.
This change removes that debug assert on the assumption that it is okay
for a peer to connect multiple times consecutively, and that the correct
behavior in that case is to just cancel the first connection and
continue as normal.
Prior to this change, the service returned by `zebra_network::init` would spawn background tasks that could silently fail, causing unexpected errors in the zebra_network service.
This change modifies the `PeerSet` that backs `zebra_network::init` to store all of the `JoinHandle`s for each background task it depends on. The `PeerSet` then checks this set of futures to see if any of them have exited with an error or a panic, and if they have it returns the error as part of `poll_ready`.
Co-authored-by: Jane Lusby <jane@zfnd.org>
Prior to this change, the seed subcommand would consistently encounter a panic in one of the background tasks, but would continue running after the panic. This is indicative of two bugs.
First, zebrad was not configured to treat panics as non recoverable and instead defaulted to the tokio defaults, which are to catch panics in tasks and return them via the join handle if available, or to print them if the join handle has been discarded. This is likely a poor fit for zebrad as an application, we do not need to maximize uptime or minimize the extent of an outage should one of our tasks / services start encountering panics. Ignoring a panic increases our risk of observing invalid state, causing all sorts of wild and bad bugs. To deal with this we've switched the default panic behavior from `unwind` to `abort`. This makes panics fail immediately and take down the entire application, regardless of where they occur, which is consistent with our treatment of misbehaving connections.
The second bug is the panic itself. This was triggered by a duplicate entry in the initial_peers set. To fix this we've switched the storage for the peers from a `Vec` to a `HashSet`, which has similar properties but guarantees uniqueness of its keys.
- Add a total peers metric to prevent races between measurements of
ready/unready peers (which can cause the sum to be wrong).
- Add an outbound request counter.
The previous outbound peer connection logic got requests to connect to new
peers and processed them one at a time, making single connection attempts
and retrying if the connection attempt failed. This was quite slow, because
many connections fail, and we have to wait for timeouts. Instead, this logic
connects to new peers concurrently (up to 50 at a time).
Attempting to implement requests for block data revealed a problem with
the previous connection logic. Block data is requested by sending a
`getdata` message with hashes of the requested blocks; the peer responds
with a sequence of `block` messages with the blocks themselves.
However, this wasn't possible to handle with the previous connection
logic, which could only convert a single Bitcoin message into a
Response. Instead, we factor out the message handling logic into a
Handler, which can statefully accumulate arbitrary data into a Response
and signal completion. This is still pretty ugly but it does work.
As a side effect, the HeartbeatNonceMismatch error is removed; because
the Handler now tries to process messages until it comes to a Response,
it just ignores mismatched nonces (and will eventually time out).
The previous Mempool and Transaction requests were removed but could be
re-added in a different form later. Also, the `Get` prefixes are
removed from `Request` to tidy the name.
Failure uses a distinct Fail trait rather than the standard library's
Error trait, which causes a lot of interoperability problems with tower
and other Error-using crates. Since failure was created, the standard
library's Error trait was improved, and its conveniences are now
available without the custom Fail trait using `thiserror` (for easy
error derives) and `anyhow` (for a better boxed Error).
* Don't expose submodules of zebra_network::peer.
* PeerSet, PeerDiscover stubs.
Co-authored-by: Deirdre Connolly <deirdre@zfnd.org>
* Initial work on PeerSet.
This is adapted from the MIT-licensed tower-balance implementation.
* Use PeerSet in the connect stub.
teor
Jane Lusby
Alfredo Garcia
Henry de Valence
Deirdre Connolly
Jane Lusby
Dimitris Apostolou