dstepanov-yadro/frostfs-node
1
0
Fork 0
forked from TrueCloudLab/frostfs-node
Code Issues Pull requests Projects Releases Packages Wiki Activity

[#1149] writecache: Prevent corruption in Head

Browse source 
Also, remove optimization comments:
1. Having to maintain an execute the same logic for headers as for
   objects is quite inefficient, as it increases memory footprint.
2. Unmarshaling object is a cheap operation if data slice is in memory.
3. For unmarshaling header-only, I think we need SDK support.

Signed-off-by: Evgenii Stratonikov <evgeniy@nspcc.ru>
This commit is contained in:
Evgenii Stratonikov 2022-03-30 11:19:30 +03:00 committed by Alex Vanin
parent ff1912aa2a
commit 48eb87d32c
1 changed files with 6 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
pkg/local_object_storage/writecache/get.go
View file

@ -16,9 +16,13 @@ func (c *cache) Get(addr *addressSDK.Address) (*objectSDK.Object, error) {
c.mtx.RLock()
for i := range c.mem {
if saddr == c.mem[i].addr {
obj := c.mem[i].obj
data := c.mem[i].data
c.mtx.RUnlock()
return obj, nil
// We unmarshal object instead of using cached value to avoid possibility
// of unintentional object corruption by caller.
// It is safe to unmarshal without mutex, as storage under `c.mem[i].data` slices is not reused.
obj := objectSDK.New()
return obj, obj.Unmarshal(data)
}
}
c.mtx.RUnlock()
@ -50,18 +54,11 @@ func (c *cache) Get(addr *addressSDK.Address) (*objectSDK.Object, error) {
//
// Returns an error of type apistatus.ObjectNotFound if requested object is missing in write-cache.
func (c *cache) Head(addr *addressSDK.Address) (*objectSDK.Object, error) {
// TODO: #1149 easiest to implement solution is presented here, consider more efficient way, e.g.:
// - provide header as common object.Object to Put, but marked to prevent correlation with full object
// (all write-cache logic will automatically spread to headers, except flushing)
// - cut header from in-memory objects directly and persist headers into particular bucket of DB
// (explicit sync with full objects is needed)
// - try to pull out binary header from binary full object (not sure if it is possible)
obj, err := c.Get(addr)
if err != nil {
return nil, err
}
// NOTE: resetting the payload via the setter can lead to data corruption of in-memory objects, but ok for others
return obj.CutPayload(), nil
}