Below is the end of a log that was never saved to S3. I was able to get it by parsing the benchmark run main instance logs on local disk. It appears that instances will not save their logs to S3 if they fail during cloud-init section. This makes debugging issues very hard.
I'd like to request that logs be saved to S3 even if the instance fails during cloud-init.
Beyond that, I'd also like to understand if the error being shown in the log can be fixed in some way. I believe it is due to spinning up many jobs at once, and by chance a few of them trigger this kind of error. Perhaps an automatic retry of some sort? Or should I switch to using Docker?
Below is the end of a log that was never saved to S3. I was able to get it by parsing the benchmark run main instance logs on local disk. It appears that instances will not save their logs to S3 if they fail during cloud-init section. This makes debugging issues very hard.
I'd like to request that logs be saved to S3 even if the instance fails during cloud-init.
Beyond that, I'd also like to understand if the error being shown in the log can be fixed in some way. I believe it is due to spinning up many jobs at once, and by chance a few of them trigger this kind of error. Perhaps an automatic retry of some sort? Or should I switch to using Docker?