ancient aspen
After reworking my contract for Preview 10 and deploying it I am getting HostError: Error(Budget, ExceededLimit) for one function on the Futurenet.
Local tests do not generate this error. Preview 9 was working fine with my function and I see no limit reduced, CPU limit was even raised.
How can I find out what limit I am hitting?
median flax
CPU limit was raised in response to re-calibrating the metering, so it's possible that you don't fit now
local tests don't generate this error because they don't enforce budget limits (probably they could, but that's not going to ever be perfect as you e.g. might run your tests with compiled contracts etc.)
also (most importantly) CPU budget (instructions) has to be specified in tx
so if you say your tx uses 1000 instructions, but it in fact uses 1001 instructions, then your tx will fail with 'budget exceeded error'
since there is no tx in tests there is really no way to alert about that earlier
ancient aspen
Thank you!
Is there any way to see number of ledger read-writes in tests?
median flax
kind of, you can call host.try_finish() to get the storage from it and then take a look at storage.footprint. every RW entry in the fooprint will be considered a ledger write
however, the setup changes you do will also end up there (like uploading wasm or creating a contract instance), so these numbers won't be very precise. the end result really depends on your transaction, so you probably should be looking at the preflight results
ancient aspen
Is preflight=simulateTransaction?
This ExceedBudget error I get from simulateTransaction and no additional info.
median flax
yeah, sorry, that's the official term for that
huh, that's weird
if you get it from simulateTransaction, then it means that there is some limit on their side and I'm not even sure if there is any... let me ask around...
median flax
if you have some unit tests, you can check the budget (IIRC it's env.budget())
spare slate
For cpu and mem limits, local tests will still enforce them against some default limits (40M cpu and 50M memory). So if your local tests pass those limits, they should likely pass the (higher) network limits too.
median flax
that's not necessarily true. if you don't deploy contracts as wasm, then you will get much smaller numbers
spare slate
Correct, assuming local tests in wasm, not native mode.
ancient aspen
Thanks guys, I will try to implement wasm local test and look into budget and RW entries.
median flax
FWIW the error you've got may only mean either running out of instructions, or running out of memory
ancient aspen
Please help me interpret the env.budget().print() results. Is 24M cpu really critical?
I have imported the contract wasm for this test and used env.budget().reset_default() right before the function in question.
=======================================================
Cpu limit: 40000000; used: 24204179
Mem limit: 52428800; used: 1695674
=======================================================
CostType cpu_insns mem_bytes
WasmInsnExec 1585493 0
WasmMemAlloc 0 0
HostMemAlloc 2867000 148697
HostMemCpy 26220 0
HostMemCmp 106351 0
InvokeHostFunction 283968 0
VisitObject 32243 0
ValXdrConv 148472 0
ValSer 11924 732
ValDeser 0 0
ComputeSha256Hash 7797 40
ComputeEd25519PubKey 0 0
MapEntry 92750 0
VecEntry 4580 0
GuardFrame 16200 1888
VerifyEd25519Sig 0 0
VmMemRead 0 0
VmMemWrite 1612 0
VmInstantiation 17854563 1543831
VmCachedInstantiation 0 0
InvokeVmFunction 5926 486
ChargeBudget 1159080 0
ComputeKeccak256Hash 0 0
ComputeEcdsaSecp256k1Key 0 0
ComputeEcdsaSecp256k1Sig 0 0
RecoverEcdsaSecp256k1Key 0 0
Int256AddSub 0 0
Int256Mul 0 0
Int256Div 0 0
Int256Pow 0 0
Int256Shift 0 0
=======================================================
After importing tokens as wasm it is much closer to the limit:
=======================================================
Cpu limit: 40000000; used: 38106611
Mem limit: 52428800; used: 3060467
=======================================================
CostType cpu_insns mem_bytes
WasmInsnExec 1901977 0
WasmMemAlloc 0 0
HostMemAlloc 2493350 163139
HostMemCpy 24288 0
HostMemCmp 135378 0
InvokeHostFunction 335008 0
VisitObject 30780 0
ValXdrConv 136412 0
ValSer 11924 732
ValDeser 0 0
ComputeSha256Hash 7797 40
ComputeEd25519PubKey 0 0
MapEntry 98315 0
VecEntry 4570 0
GuardFrame 16200 1888
VerifyEd25519Sig 0 0
VmMemRead 0 0
VmMemWrite 1736 0
VmInstantiation 31698792 2892724
VmCachedInstantiation 0 0
InvokeVmFunction 23704 1944
ChargeBudget 1186380 0
ComputeKeccak256Hash 0 0
ComputeEcdsaSecp256k1Key 0 0
ComputeEcdsaSecp256k1Sig 0 0
RecoverEcdsaSecp256k1Key 0 0
Int256AddSub 0 0
Int256Mul 0 0
Int256Div 0 0
Int256Pow 0 0
Int256Shift 0 0
=======================================================
FWIW futurenet limit is 100M, so seems well below that...
but this doesn't really matter: the error you're getting happens at runtime. if you tried to submit a transaction that declares that it uses 101M instructions, we would reject it immediately. but since your contract fails at apply time, that means that it uses more instructions than you've declared in your transaction, not more instructions than network limit
i.e. you should compare these numbers to the numbers you have in your transaction's sorobanData
ancient aspen
i.e. you should compare these numbers to the numbers you have in your transactio...
is there such data in a transaction for simulateTransaction?
it is failing on simulating part when I try to determine a fee.
median flax
ah, sorry, forgot about that... that's suspicious indeed
median flax
just double-checked that for preflight the CPU limit is 40M (definitely should be bumped to match the network limit...) given that you are pretty close, I guess you might be just reaching it currently? there might be some discrepancy in the setup/enivonment between preflight and local test and it's really just 2% more compared to your numbers
ancient aspen
I see, that most likely is blocking my transaction. I will try set fee manually and not use simulateTransaction, will it help to avoid preflight?
median flax
you'll need to build the proper transaction yourself, which might be a bit involved depending on how complex it is (you'll need to fill in the footprint and resources manually). you could maybe try optimizing the contracts a bit (e.g. run them through wasm-opt to reduce the contract size and thus vm instantiation cost)
there is also a CLI command for that IIRC
ancient aspen
Thank you! I'll try. Hope limit will be raised to 100M soon.
median flax
yeah, we'll try to fix this. thanks for pointing this out and sorry for all the time it took
ancient aspen
@median flax
Hi, is a changed preflight CPU limit on RPC in effect yet?
I am still hitting the limit after all optimizations and wonder if it could be another limit, not CPU?
median flax
I don't think this has been released yet...
olive ocean
We're running into the same.
olive ocean
I'm trying to understand the various numbers here. Running against a WASM the failing function gives something very similar to @ancient aspen , 38M against 40M limit.
Now futurenet has 100M but it fails and I understand that currently the preflight / simulate transaction has still 40M but may function fails at prepareTransaction - is that the same thing?
Ah i see in the source that it's the same thing.
tranquil turret
Hi, has the change been released here? @median flax
median flax
unfortunately we won't be able to update the preview 10 release, so the fix will only be in the next release
tranquil turret
@median flax when is that next release expected to launch?
median flax
when it's ready - but probably within a few weeks
tranquil turret
Hey @median flax -- wanted to follow up here to see if there is any more clarity regarding the next release?
median flax
I'm not sure I can share the exact date and it's definitely subject to change. it's still within few weeks (so not tomorrow, but also probably not in 2 months)
simple yacht
Slender running into the same for simple tx with just read from another contract. It consumes 10122617 cpu_insns in tests metering, but fails on prefligth with ExceedBudget in Futurenet
olive ocean
Is that fixed now?
median flax
yes, preflight limits are now consistent with the futurenet limits
median flax
yes, sure