How do I rerun something until succeed in the background with pipe in linux command line? - linux

Question

Story: I'm following this guide to setup my geth node:
https://github.com/enthusiastics/bsc-archive-snapshot/blob/master/build_archive_node.sh
However, for this code chunk
# Query S3 for all archives and download them in parallel to a new zfs dataset.
while IFS= read -r FILE_NAME; do
ZFS_NAME=$(echo "$FILE_NAME" | cut -d'.' -f1)
ARCHIVE_NAMES+=("$ZFS_NAME")
zfs create -o "mountpoint=/$ZFS_NAME" "tank/$ZFS_NAME"
bash -c "cd /$ZFS_NAME && aws s3 cp --request-payer=requester '$S3_BUCKET_PATH/bsc/$FILE_NAME' - | /zstd/zstd --long=30 -d | tar -xf -" &
done <<<"$(aws s3 ls --request-payer=requester "$S3_BUCKET_PATH/bsc/" | cut -d' ' -f4)"
To be exact for the following line
bash -c "cd /$ZFS_NAME && aws s3 cp --request-payer=requester '$S3_BUCKET_PATH/bsc/$FILE_NAME' - | /zstd/zstd --long=30 -d | tar -xf -" &
some of instance would results in broken pipe, but a automatic restart could help. So I want to rewrite the code so that it automatically retries until the above line is succeeded. I researched a bit and found the until do done method. Example
until passwd ; do echo "Try again" ; done;
However, I could not succeed in incorporating the above idea into the GitHub code. I try to rewrite it as:
bash -c "cd /$ZFS_NAME && until aws s3 cp --request-payer=requester '$S3_BUCKET_PATH/bsc/$FILE_NAME' - | /zstd/zstd --long=30 -d | tar -xf - ; do echo "Try again" ; done;" &
But it does not work... I test it on the minimum code:
until passwd | echo "randompipe" ; do echo "Try again" ; done;
it didn't run passwd until succeed