cerc-io/lotus
10
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 27 Wiki Activity

Reduce LOTUS_FMV_CONCURRENCY

Browse Source 
Reduce LOTUS_FMV_CONCURRENCY
This commit is contained in:
Phi 2023-04-21 22:02:33 +02:00
parent 19d8f79a3d
commit b7484121fc
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
CHANGELOG.md
View File

@ -13,7 +13,7 @@ Please read carefully through the **upgrade warnings** section if you are upgrad
- **Storage Providers:** The proofs libraries now have CUDA enabled by default, which requires you to install (CUDA)[https://lotus.filecoin.io/tutorials/lotus-miner/cuda/] if you haven't already done so. If you prefer to use OpenCL on your GPUs instead, you can use the `FFI_USE_OPENCL=1` flag when building from source. On the other hand, if you want to disable GPUs altogether, you can use the `FFI_NO_GPU=1` environment variable when building from source.
- **Storage Providers:** The `lotus-miner sectors extend` command has been refactored to the functionality of `lotus-miner sectors renew`.
- **Exchanges/Node operators/RPC-providers::** Execution traces (returned from `lotus state exec-trace`, `lotus state replay`, etc.), has changed to account for changes introduced by the by the FVM. **Please make sure to read the `Execution trace format change` section carefully, as these are interface breaking changes**
- **Syncing issues:** If you have been struggling with syncing issues in normal operations you can try to adjust the amount of threads used for more concurrent FMV execution through via the `LOTUS_FVM_CONCURRENCY` enviroment variable. It is set to 4 threads by default. Recommended formula for concurrency == YOUR_RAM/4 , but max is 128. If you are a Storage Provider and are pushing many messages within a short period of time, exporting `LOTUS_SKIP_APPLY_TS_MESSAGE_CALL_WITH_GAS=1` will also help with keeping in sync.
- **Syncing issues:** If you have been struggling with syncing issues in normal operations you can try to adjust the amount of threads used for more concurrent FMV execution through via the `LOTUS_FVM_CONCURRENCY` enviroment variable. It is set to 4 threads by default. Recommended formula for concurrency == YOUR_RAM/4 , but max during a network upgrade is 24. If you are a Storage Provider and are pushing many messages within a short period of time, exporting `LOTUS_SKIP_APPLY_TS_MESSAGE_CALL_WITH_GAS=1` will also help with keeping in sync.
- **Catching up from a Snapshot:** Users have noticed that catching up sync from a snapshot is taking a lot longer these day. This is largely related to the built-in market actor consuming a lot of computational demand for block validation. A FIP for a short-term mitigation for this is currently in Last Call and will be included network version 19 upgrade if accepted. You [can read the FIP here.](https://github.com/filecoin-project/FIPs/blob/master/FIPS/fip-0060.md)
## Highlights