Hi,
I'm using diskpacked storage engine, as I don't want to pay the price of lot of small files.
Blobpacked came, and I've measured that is quite good - and with the new "union" storage, I can even move its localdisk files into a diskpacked store.
But blobpacked packs blobs of the same file, so the blobs should be copied into the blobpacked store in "natural" order, not lexicographical, as sync does.
That's what BlobStreamer interface is for. But that interface does not have a HTTP endpoint, as BlobEnumerator (sync) has.
I've tried to implement it, but it is a big undertaking, and I only want to shovel my blobs from my diskpacked into my blobpacked.
The I thought maybe camput could get a "rawdpstream" subcommand, and "cat diskpacked/pack-*.blobs | camput rawdpstream",
but camput is too intelligent, and I cannot specify the destination exactly - it only accepts the server as the server prefix,
contrary to camtool sync, which can.
Shall I try to add a "stream-diskpacked" subcommand to camtool?
Or any other idea?
Thanks,
Tamás Gulácsi