Large downloads keep dying? Resume and multi-source patterns that work
Large-file failures usually mix three causes (network, server, local disk) and treat them as one.
If you only want the conclusion, the last section is self-contained.
Classify before you retry
Reset connections / progressive slowdown — throttling or inter-peak congestion. Different mirror or a different hour.
Stalls at an exact offset — proxy timeout. Raise it or bypass the proxy for this host.
Complete-looking file, failed digest — the write was truncated. Delete it; never attempt to resume a file that already failed verification.
When resume actually resumes
curl -C - works only when the server honors Range and the local length is correct:
curl -C - -L -o pkg.tar.bz2 https://example.invalid/pkg.tar.bz2
wget -c attempts resume by default: wget -c pkg.tar.bz2.
The resumed file still has to produce the official SHA-256. Resume saves bandwidth; it never waives verification.
Multi-source has a hard boundary
Parallel connections to the same URL are fine:
aria2c -x 8 -s 8 -k 1M pkg.iso
Splicing bytes from different mirrors is not. Two mirrors can host different builds; interleaved chunks produce a file that never matches any digest.
Rule: parallelize one URL, fall back across URLs — never mix across URLs within one file.
Related resource
For the "can I re-verify this later" half of the workflow I keep GitVaults (https://gitvaults.com) open: every entry carries the upstream link, file size, and SHA-256, with a cloud mirror as the reachability fallback.
metasequoiaime msime windows 0 9 1 → https://gitvaults.com/download/2026/09/28/metasequoiaime-msime-windows-0-9-1/en
That page puts version, architecture, and digest on one screen — download, compare the hash, then install. The catalog has another hundred-plus entries in the same format if you need neighbors.
Caveat
This is what works in our environment; your constraints may differ. If you have a sturdier sequence, corrections are welcome in the comments.
Top comments (0)