Liquidator: add Sanctum swap #919
Conversation
farnyser
force-pushed
the
serge/lst-swap
branch
4 times, most recently
from
5723575 to
66bdbc8
Compare
farnyser
changed the title
| } | ||
| } | ||
|
|
||
| let results = futures::future::join_all(jobs).await; |
There was a problem hiding this comment.
future pr: instead of configuring individual timeouts, it could be good to give a general quote timeout that applies here
bin/liquidator/src/rebalance.rs
Outdated
| let accounts: Vec<Account> = live_rpc_client | ||
| .get_multiple_accounts(&lookup_table.addresses) | ||
| .await? | ||
| .drain(..) |
There was a problem hiding this comment.
wouldn't iter() do the same thing?
There was a problem hiding this comment.
iter() give a &Option<T> and I want a Option<T>
| let in_price = get_or_fetch_price(r.input_mint); | ||
| let out_price = get_or_fetch_price(r.output_mint); | ||
| let amount = out_price * I80F48::from_num(r.out_amount) | ||
| - in_price * I80F48::from_num(r.in_amount); |
There was a problem hiding this comment.
mmh, this is not a sufficient criterion now that you've removed the favoring of the primary route
say someone wants to swap mSOL -> KIN. Due to the configuration we also check the alternate route mSOL -> SOL. This latter route is near guaranteed to have better amount than the one to KIN. But it will still often do worse if you factor in the SOL -> KIN second step that's needed afterwards.
suggestion: two step approach?
- fetch the routes between the correct in and out mints first (including sanctum if enabled and it's both LSTs) and find the best route (by
amount) that fits in a transaction - if this fails, fetch the alternate routes (which are just the first step of a multi-part route) and grab the best one of those the same way
that also reduces the number of quote requests sent to the swap endpoints (being rate limited is a real concern)
bin/liquidator/src/rebalance.rs
Outdated
| } | ||
| } | ||
|
|
||
| async fn load_lst(&mut self, live_rpc_client: &RpcClient) -> anyhow::Result<HashSet<Pubkey>> { |
There was a problem hiding this comment.
this does not even need to take a self argument anymore
bin/liquidator/src/rebalance.rs
Outdated
| let accounts: Vec<Account> = live_rpc_client | ||
| .get_multiple_accounts(&lookup_table.addresses) | ||
| .await? | ||
| .drain(..) |
lib/client/src/gpa.rs
Outdated
| .await; | ||
|
|
||
| match req_res { | ||
| Ok(v) => Ok(keys.drain(..).zip(v.value).collect::<Vec<_>>()), |
There was a problem hiding this comment.
really think the .drain(..) looks odd, into_iter() again?
There was a problem hiding this comment.
minor: this possibly got overlooked
lib/client/src/gpa.rs
Outdated
| .drain(..) | ||
| .map(|v| v.unwrap()) | ||
| .flatten() | ||
| .filter_map(|x| x.1.map(|o| (x.0, o))) |
There was a problem hiding this comment.
prefer: .filter_map(|(pubkey, account_opt)| account_opt.map(|acc| (pubkey, acc))) (but even that needs some thought to read, unfortunately)
Is it even acceptable to not return data for some of the accounts that were requested? Could you put in the doc comment that this is could return no data without an error?
liquidator: add sanctum swap (cherry picked from commit 01d5237)
* liquidator: randomly select token/perps for rebalancing to avoid failing at every try if one token is having an issue (#921) (cherry picked from commit e3a7ed9) * liquidator: do not panic if token or perp rebalancing fails (#927) (cherry picked from commit 0b7e62e) * Liquidator: add Sanctum swap (#919) liquidator: add sanctum swap (cherry picked from commit 01d5237) * liquidator: add more LST for sanctum swap (#944) (cherry picked from commit c0b61b3)
No description provided.