curvefs: a curve filesystem can be mounted by multi fuse clients #1101
Conversation
ilixiaocui
force-pushed
the
master
branch
2 times, most recently
from
5b14bd9 to
3be190c
Compare
curvefs/src/client/fuse_client.h
Outdated
|
|
||
| virtual ~FuseClient() {} | ||
|
|
||
| FuseClient(const std::shared_ptr<MdsClient> &mdsClient, | ||
| const std::shared_ptr<MetaServerClient> &metaClient, |
There was a problem hiding this comment.
Is it necessary to change the format？there are many similar
| VLOG(1) << "update inode open status"; | ||
| int32_t oldcount = old->openmpcount(); | ||
| int32_t newcount = | ||
| request.inodeopenstatuschange() == InodeOpenStatusChange::OPEN |
There was a problem hiding this comment.
I think the status judgment of InodeOpenStatusChange is a bit confusing here, directly judge open count+1, close count-1
There was a problem hiding this comment.
According to the request. Inodeopenstatuschange ()， openmpcount plus or minus one.
Here is conditional operator, I use it instead of if else because it only has two conditions.
Exp1 ? Exp2 : Exp3;
curvefs/src/client/inode_wrapper.h
Outdated
| // TODO(xuchaojie) : get from conf maybe? | ||
| const uint32_t kOptimalIOBlockSize = 0x10000u; | ||
|
|
||
| using rpcclient::MetaServerClient; | ||
| using rpcclient::MetaServerClientImpl; | ||
|
|
||
| std::ostream &operator<<(std::ostream &os, const struct stat &attr); | ||
| void AppendS3ChunkInfoToMap(uint64_t chunkIndex, const S3ChunkInfo &info, | ||
| void AppendS3ChunkInfoToMap( |
There was a problem hiding this comment.
Too much formatting, almost nowhere to find actually modify place...
| enum InodeOpenStatusChange { | ||
| OPEN = 1; | ||
| CLOSE = 2; | ||
| NOCHANGE = 3; |
There was a problem hiding this comment.
MetaStatusCode UpdateInode(const Inode &inode,
InodeOpenStatusChange statusChange =
InodeOpenStatusChange::NOCHANGE) = 0;
UpdateInode adds a parameter InodeOpenStatusChange to update the InodeOpenStatusChange in UpdateInodeRequest，NOCHANGE is the default parameter.
ilixiaocui
force-pushed
the
master
branch
3 times, most recently
from
f75daad to
9a21f11
Compare
NOTES: Now we implemented close-to-open consistency and multi-mountpoint for one filesystem. A file written by a client, closed, and then opened by any client can read the latest data. The scenario where multiple mount points modify the same resource ( Any modification to a file ) at the same time is not guaranteed.
NOTES:
Now we implemented close-to-open consistency and multi-mountpoint for one filesystem.
A file written by a client, closed, and then opened by any client can read the latest data.
The scenario where multiple mount points modify the same resource ( Any modification to a
file ) at the same time is not guaranteed.
What problem does this PR solve?
Issue Number: close #xxx
Problem Summary:
What is changed and how it works?
What's Changed:
How it Works:
Side effects(Breaking backward compatibility? Performance regression?):
Check List