mirror of
https://github.com/seaweedfs/seaweedfs.git
synced 2024-12-18 12:47:50 +08:00
1a194a578c
Among the changes, this replaces a couple instances of "Seaweed File System" with "SeaweedFS", for the same reason that nobody says "Mongo Data Base".
47 lines
1.3 KiB
Plaintext
47 lines
1.3 KiB
Plaintext
How to submit a content
|
|
1. Find physical volumes
|
|
1.c Create a hash value
|
|
1.d find a write logic volume id, and return [logic volume id, {physical volume ids}]
|
|
2. submit to physical volumes
|
|
2.c
|
|
generate the cookie
|
|
generate a unique id as key
|
|
choose the right altKey
|
|
send bytes to physical volumes
|
|
2.s each
|
|
save bytes
|
|
store map[key uint64, altKey uint32]<offset, size>
|
|
for updated entry, set old entry's offset to zero
|
|
3.c
|
|
wait for all physical volumes to finish
|
|
store the /<logic volume id>/<key>_<cookie>_<altKey>.<ext>
|
|
|
|
How to retrieve a content
|
|
1.c
|
|
send logic volume id
|
|
1.d
|
|
find least busy volume's id
|
|
2.c
|
|
send URI /<physical volume id>/<key>_<cookie>_<altKey>.<ext>
|
|
|
|
|
|
How to submit a content
|
|
1. send bytes to SeaweedFS, got <volume id, key uint64, cookie code>
|
|
store <key uint64, volume id uint32, cookie code uint32, ext>, and other information
|
|
|
|
To read a content
|
|
2. use logic volume id to lookup a <machine id>
|
|
render url as /<machine id>/<volume id>/<key>/<cookie>.ext
|
|
|
|
The directory server
|
|
0.init
|
|
load and collect <logic volume id, machine ids> mapping
|
|
1.on submit content
|
|
find a free logic volume id, start sending content to 3 machines
|
|
if all of them finishes, return <logic volume id, key, cookie code>
|
|
2.on read content
|
|
based on logic volume id, pick a machine with less load,
|
|
return <machine id>
|
|
|
|
|