-
Notifications
You must be signed in to change notification settings - Fork 44
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
support send large file to workload #532
Conversation
cluster/cluster.go
Outdated
@@ -73,6 +73,7 @@ type Cluster interface { | |||
// file methods | |||
Copy(ctx context.Context, opts *types.CopyOptions) (chan *types.CopyMessage, error) | |||
Send(ctx context.Context, opts *types.SendOptions) (chan *types.SendMessage, error) | |||
SendLargeFile(ctx context.Context, opts chan *types.SendLargeFileOptions, resp chan *types.SendMessage) error |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
其实不用加新的 interface, 直接改旧的 Send 就可以了.
要是有问题可以开个 zoom 来讨论 |
我改了下代码,现在看起来是能够跑了。但是会有些比较诡异的情况,时而成功时而失败,让我再debug一下。然后关于上面的review意见我也再改一下。 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
自己多测试一下, 包括老的 Send 接口是否还能用, 还有发送一些 >3G 的文件, 都是我们真实存在的业务场景.
cluster/calcium/sendlarge.go
Outdated
utils.SentryGo(func(ID, name string, size int64, content io.Reader, uid, gid int, mode int64) func() { | ||
return func() { | ||
defer wg.Done() | ||
if err := c.withWorkloadLocked(ctx, ID, func(ctx context.Context, workload *types.Workload) error { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
每个 chunk 都上一次锁, 虽然临界区是小了, 但是代价可能有点大. 考虑下从头到尾只上一把大锁, 传个大文件(>1G)对比一下耗时.
@@ -39,6 +39,7 @@ type API interface { | |||
VirtualizationCreate(ctx context.Context, opts *enginetypes.VirtualizationCreateOptions) (*enginetypes.VirtualizationCreated, error) | |||
VirtualizationResourceRemap(context.Context, *enginetypes.VirtualizationRemapOptions) (<-chan enginetypes.VirtualizationRemapMessage, error) | |||
VirtualizationCopyTo(ctx context.Context, ID, target string, content []byte, uid, gid int, mode int64) error | |||
VirtualizationCopyChunkTo(ctx context.Context, ID, target string, size int64, content io.Reader, uid, gid int, mode int64) error |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
CopyTo 和 CopyChunkTo 可以合并吧, 让 cluster.Copy 和 cluster.CopyLarge 都调用一个 engine 接口
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
刚才简单讨论了一下, rpc 层还是两个 API, 但是 cluster 层和 engine 都只有一个 send 接口.
06977aa
to
837bec8
Compare
0c06f3a
to
34dacc4
Compare
unbelievable |
a982876
to
924a9a8
Compare
resolve conflicts. |
另外注意一下单元测试覆盖率和实际测试下多路文件并发 send |
关于cli的send命令,之前的
send
是直接整个文件发到core里面,但是core的grpc配置为单条message最大只能为20M,所以有些文件是无法传输到core端的。这里是增加了一个叫做
SendLargeFile
的方法,用stream
的方式接受cli端分chunk
发送过来的数据,并拼接起来再一并发送到容器里面。