diff --git a/docs/research/failure-handling.md b/docs/research/failure-handling.md
index 56692ad..c0108b5 100644
--- a/docs/research/failure-handling.md
+++ b/docs/research/failure-handling.md
@@ -10,7 +10,7 @@ Generally we should allow user to provide desired reaction of AppController on p
 Kubernetes object. Possible reactions include:
 
  1. **Retry**: AC tries to delete failed K8s resource and then re-create it. We should be able to set maximum retries and delay before the next retry.
- 2. **Rollback**: AC deletes all resourses that were created during current run.
+ 2. **Rollback**: AC deletes all resources that were created during current run.
  3. **Abort**: Immediate quit, keep all created resources.
  4. **Ignore current**: AC proceeds with creating resources as if current one was created successfully.
  5. **Ignore children**: AC skips creation of all resources depending on the current one and all cascade dependencies, but proceeds with creating the rest of resources.