[curve/toos-v2]: add snapshot-clone-status #2346

Cyber-SiKu · 2023-03-30T06:52:28Z

Is your feature request related to a problem? (你需要的功能是否与某个问题有关?)

we'd like to support bs status snapshot command in curve tool

The implementation of the old tool is here：

Lines 658 to 685 in 8d6caf7

    
           int StatusTool::PrintSnapshotCloneStatus() { 
        
               std::cout << "SnapshotCloneServer status:" << std::endl; 
        
               if (noSnapshotServer_) { 
        
                   std::cout << "No SnapshotCloneServer" << std::endl; 
        
                   return 0; 
        
               } 
        
               std::string version; 
        
               std::vector<std::string> failedList; 
        
               int res = versionTool_->GetAndCheckSnapshotCloneVersion(&version, 
        
                                                                       &failedList); 
        
               int ret = 0; 
        
               if (res != 0) { 
        
                   std::cout << "GetAndCheckSnapshotCloneVersion fail" << std::endl; 
        
                   ret = -1; 
        
               } else { 
        
                   std::cout << "version: " << version << std::endl; 
        
                   if (!failedList.empty()) { 
        
                       versionTool_->PrintFailedList(failedList); 
        
                       ret = -1; 
        
                   } 
        
               } 
        
               std::vector<std::string> activeAddrs = snapshotClient_->GetActiveAddrs(); 
        
               std::map<std::string, bool> onlineStatus; 
        
               snapshotClient_->GetOnlineStatus(&onlineStatus); 
        
               std::cout << "current snapshot-clone-server: " << activeAddrs << std::endl; 
        
               PrintOnlineStatus("snapshot-clone-server", onlineStatus); 
        
               return ret; 
        
           }

The old command input and out put:

curve_ops_tool snapshot-clone-status

-------
output:
SnapshotCloneServer status:
version: ***
current snapshot-clone-server: ***
online snapshot-clone-server list: ***
offline snapshot-clone-server list: ***

Refer to tool develop guide to get start, and paste the result of the command in pr.

Build compilation environment：https://github.com/opencurve/curve/blob/master/docs/cn/build_and_run.md

Describe the solution you'd like (描述你期望的解决方法)

Add subcommand clinet to curve bs list.

Describe alternatives you've considered (描述你想到的折衷方案)

Additional context/screenshots (更多上下文/截图)

The text was updated successfully, but these errors were encountered:

zhanghuidinah · 2023-04-12T06:30:46Z

This issue has been chosen for community developer activities. Welcome to feel free to try it. After the PR is merged, you will receive a gift package and a certificate as a community contributor.
Developer Activities Registration Steps please ref [2023 Q2 Developer Activities]: Call For Participation！ #2334 (comment)

Xinlong-Chen · 2023-04-21T11:35:49Z

@Cyber-SiKu I am interested in this task, please assign it to me, thanks.

ilixiaocui · 2023-04-23T01:41:37Z

Xinlong-Chen

welcome!

Xinlong-Chen · 2023-04-25T03:27:43Z

@Cyber-SiKu When I tried to understand this task, I found that this task is similar to the implementation of the curve bs status mds command.

But I have a little doubt about the implementation of mds.

In tools-v1: there need 3 rpcs to get status(in curve/src/tools/status_tool.cpp@PrintMdsStatus):

versionTool_->GetAndCheckMdsVersion(&version, &failedList)
mdsClient_->GetCurrentMds()
mdsClient_->GetMdsOnlineStatus(&onlineStatus)

But this is different in tools-v2.

There are only 2 rpcs in tools-v2(in curve/tools-v2/pkg/cli/command/curvebs/status/mds/mds.go@Init):

statusMetric := basecmd.NewMetric(addrs, STATUS_SUBURI, timeout)
versionMetric := basecmd.NewMetric(addrs, VERSION_SUBURI, timeout)

This 2 rpcs in tools-v2 is the same as the first two rpcs in tools-v1.

I would like to ask why it can be done like this (get rid of the tail rpc)?

Cyber-SiKu · 2023-04-25T04:46:54Z

@Cyber-SiKu When I tried to understand this task, I found that this task is similar to the implementation of the curve bs status mds command.

But I have a little doubt about the implementation of mds.

In tools-v1: there need 3 rpcs to get status(in curve/src/tools/status_tool.cpp@PrintMdsStatus):

versionTool_->GetAndCheckMdsVersion(&version, &failedList)

mdsClient_->GetCurrentMds()

mdsClient_->GetMdsOnlineStatus(&onlineStatus)

But this is different in tools-v2.

There are only 2 rpcs in tools-v2(in curve/tools-v2/pkg/cli/command/curvebs/status/mds/mds.go@Init):

statusMetric := basecmd.NewMetric(addrs, STATUS_SUBURI, timeout)

versionMetric := basecmd.NewMetric(addrs, VERSION_SUBURI, timeout)

This 2 rpcs in tools-v2 is the same as the first two rpcs in tools-v1.

I would like to ask why it can be done like this (get rid of the tail rpc)?

GetCurrentMds

curve/src/tools/mds_client.cpp

Lines 973 to 990 in 4e73df5

    
           std::vector<std::string> MDSClient::GetCurrentMds() { 
        
               std::vector<std::string> leaderAddrs; 
        
               for (const auto item : dummyServerMap_) { 
        
                   // 获取status来判断正在服务的地址 
        
                   std::string status; 
        
                   MetricRet ret = metricClient_.GetMetric(item.second, 
        
                                                   kMdsStatusMetricName, &status); 
        
                   if (ret != MetricRet::kOK) { 
        
                       std::cout << "Get status metric from " << item.second 
        
                                 << " fail" << std::endl; 
        
                       continue; 
        
                   } 
        
                   if (status == kMdsStatusLeader) { 
        
                       leaderAddrs.emplace_back(item.first); 
        
                   } 
        
               } 
        
               return leaderAddrs; 
        
           }

GetMdsOnlineStatus

curve/src/tools/mds_client.cpp

Lines 901 to 925 in 4e73df5

    
           int MDSClient::GetListenAddrFromDummyPort(const std::string& dummyAddr, 
        
                                                     std::string* listenAddr) { 
        
               assert(listenAddr != nullptr); 
        
               MetricRet res = metricClient_.GetConfValueFromMetric(dummyAddr, 
        
                                   kMdsListenAddrMetricName, listenAddr); 
        
               if (res != MetricRet::kOK) { 
        
                   return -1; 
        
               } 
        
               return 0; 
        
           } 
        
           void MDSClient::GetMdsOnlineStatus(std::map<std::string, bool>* onlineStatus) { 
        
               assert(onlineStatus != nullptr); 
        
               onlineStatus->clear(); 
        
               for (const auto item : dummyServerMap_) { 
        
                   std::string listenAddr; 
        
                   int res = GetListenAddrFromDummyPort(item.second, &listenAddr); 
        
                   // 如果获取到的监听地址与记录的mds地址不一致，也认为不在线 
        
                   if (res != 0 || listenAddr != item.first) { 
        
                       onlineStatus->emplace(item.first, false); 
        
                       continue; 
        
                   } 
        
                   onlineStatus->emplace(item.first, true); 
        
               } 
        
           }

It can be found that both are essentially obtaining metirc to judge the status of the server.It's just that v1 still needs to judge whether the dummyport is consistent with the configuration file.

Xinlong-Chen · 2023-04-25T06:18:52Z

Got it!

Based on forementioned reasons, we also can implement bs status snapshot command with 2 rpcs.

There are the corresponding metric strings I found about snapshot. I don't know if that's right?

STATUS_SUBURI  = "/vars/snapshotcloneserver_status"
VERSION_SUBURI = "/vars/curve_version"

Cyber-SiKu · 2023-04-25T06:40:16Z

Got it!

Based on forementioned reasons, we also can implement bs status snapshot command with 2 rpcs.

There are the corresponding metric strings I found about snapshot. I don't know if that's right?
STATUS_SUBURI  = "/vars/snapshotcloneserver_status"
VERSION_SUBURI = "/vars/curve_version"

It should be like this, you can refer to the old code. It seems that there are differences in versions

Xinlong-Chen · 2023-04-25T11:46:12Z

Well, I also have a question that how can I test it?
When I use tools-v1 command curve_ops_tool snapshot-clone-status in docker(use reference image)，the following result I get bothers me.

root@f8df310173eb:/# curve_ops_tool snapshot-clone-status
SnapshotCloneServer status:
No SnapshotCloneServer

Do I need do more action about snapshot deployment to get correct result as following:

curve_ops_tool snapshot-clone-status

-------
output:
SnapshotCloneServer status:
version: ***
current snapshot-clone-server: ***
online snapshot-clone-server list: ***
offline snapshot-clone-server list: ***

Cyber-SiKu · 2023-04-25T12:44:15Z

Currently snapshot-related commands cannot use the cluster deployed by curveadm palygroud (there is no s3, so there is no snapshot).
Need to deploy a minio.
Then follow the deployment document to deploy a stand-alone cluster. You can skip formatting the disk, and then fill in the configuration items of topology.yaml:

kind: curvebs
global:
...
   s3.nos_address: <> // ip:9000 is the ip and port number deployed by minio
   s3.snapshot_bucket_name: <> // created bucket name
   s3.ak: <> // ak minioadmin
   s3.sk: <> //sk minioadmin
...

chunkserver_services:
   config:
...
     copiesets: 100
      chunkfilepool.enable_get_chunk_from_pool: false
   deploy:
     - host: ${target}
     - host: ${target}
     - host: ${target}
...

Cyber-SiKu · 2023-04-25T12:44:23Z

目前snapshot 相关的命令不能使用 curveadm palygroud 部署的集群(没有s3,所以没有snapshot).
需要先部署一个minio.
然后按照部署文档部署一个单机集群.可以跳过格式化磁盘,然后在填写topology.yaml的配置项:

kind: curvebs
global:
... ...
  s3.nos_address: <> // ip:9000 为minio部署的ip和端口号
  s3.snapshot_bucket_name: <> // 创建的桶名
  s3.ak: <> // ak  minioadmin
  s3.sk: <> //sk minioadmin
... ...

chunkserver_services:
  config:
... ...
    copysets: 100
     chunkfilepool.enable_get_chunk_from_pool: false
  deploy:
    - host: ${target}
    - host: ${target}
    - host: ${target}
... ...

Xinlong-Chen · 2023-04-25T16:25:20Z

It seems that both addr and dummyaddr should be given by user(or file).
in tools-v1(mds):

curve/src/tools/mds_client.cpp

Lines 37 to 77 in 4e73df5

    
           int MDSClient::Init(const std::string& mdsAddr, 
        
                               const std::string& dummyPort) { 
        
               if (isInited_) { 
        
                   return 0; 
        
               } 
        
               // 初始化channel 
        
               curve::common::SplitString(mdsAddr, ",", &mdsAddrVec_); 
        
               if (mdsAddrVec_.empty()) { 
        
                   std::cout << "Split mds address fail!" << std::endl; 
        
                   return -1; 
        
               } 
        
               int res = InitDummyServerMap(dummyPort); 
        
               if (res != 0) { 
        
                   std::cout << "init dummy server map fail!" << std::endl; 
        
                   return -1; 
        
               } 
        
               for (uint64_t i = 0; i < mdsAddrVec_.size(); ++i) { 
        
                   if (channel_.Init(mdsAddrVec_[i].c_str(), nullptr) != 0) { 
        
                       std::cout << "Init channel to " << mdsAddr << "fail!" << std::endl; 
        
                       continue; 
        
                   } 
        
                   // 寻找哪个mds存活 
        
                   curve::mds::topology::ListPhysicalPoolRequest request; 
        
                   curve::mds::topology::ListPhysicalPoolResponse response; 
        
                   curve::mds::topology::TopologyService_Stub stub(&channel_); 
        
                   brpc::Controller cntl; 
        
                   cntl.set_timeout_ms(FLAGS_rpcTimeout); 
        
                   stub.ListPhysicalPool(&cntl, &request, &response, nullptr); 
        
                   if (cntl.Failed()) { 
        
                       continue; 
        
                   } 
        
                   currentMdsIndex_ = i; 
        
                   isInited_ = true; 
        
                   return 0; 
        
               } 
        
               std::cout << "Init channel to all mds fail!" << std::endl; 
        
               return -1; 
        
           }

in tools-v1(snapshot):

curve/src/tools/snapshot_clone_client.cpp

Lines 28 to 42 in 4e73df5

    
           int SnapshotCloneClient::Init(const std::string& serverAddr, 
        
                                         const std::string& dummyPort) { 
        
               curve::common::SplitString(serverAddr, ",", &serverAddrVec_); 
        
               if (serverAddrVec_.empty()) { 
        
                   // no snapshot clone server 
        
                   return 1; 
        
               } 
        
               int res = InitDummyServerMap(dummyPort); 
        
               if (res != 0) { 
        
                   std::cout << "init dummy server map fail!" << std::endl; 
        
                   return -1; 
        
               } 
        
               return 0; 
        
           }

in tools-v2(mds):

curve/tools-v2/pkg/config/bs.go

Lines 195 to 211 in 4e73df5

    
           func AddBsUint64RequiredFlag(cmd *cobra.Command, name string, usage string) { 
        
           	cmd.Flags().Uint64(name, uint64(0), usage+color.Red.Sprint("[required]")) 
        
           	cmd.MarkFlagRequired(name) 
        
           	err := viper.BindPFlag(BSFLAG2VIPER[name], cmd.Flags().Lookup(name)) 
        
           	if err != nil { 
        
           		cobra.CheckErr(err) 
        
           	} 
        
           } 
        
           // add flag option 
        
           // bs mds[option] 
        
           func AddBsMdsFlagOption(cmd *cobra.Command) { 
        
           	AddBsStringOptionFlag(cmd, CURVEBS_MDSADDR, "mds address, should be like 127.0.0.1:6700,127.0.0.1:6701,127.0.0.1:6702") 
        
           } 
        
           func AddBsMdsDummyFlagOption(cmd *cobra.Command) { 
        
           	AddBsStringOptionFlag(cmd, CURVEBS_MDSDUMMYADDR, "mds dummy address, should be like 127.0.0.1:6700,127.0.0.1:6701,127.0.0.1:6702") 
        
           }

I guess that the strings is from user input or config file like template.yaml:

curve/tools-v2/pkg/config/template.yaml

Lines 21 to 23 in 4e73df5

    
           mdsAddr: 127.0.0.1:6700,127.0.0.1:6701,127.0.0.1:6702 
        
           mdsDummyAddr: 127.0.0.1:7700,127.0.0.1:7701,127.0.0.1:7702 
        
           etcdAddr: 127.0.0.1:23790,127.0.0.1:23791, 127.0.0.1:23792

How can i get the addr and dummy in snapshot tools-v2 implement?(add two kv pairs about snapshot in template.yaml?)

Cyber-SiKu · 2023-04-26T03:24:00Z

It seems that both addr and dummyaddr should be given by user(or file). in tools-v1(mds):

curve/src/tools/mds_client.cpp

Lines 37 to 77 in 4e73df5

int MDSClient::Init(const std::string& mdsAddr,

const std::string& dummyPort) {

if (isInited_) {

return 0;

}

// 初始化channel

curve::common::SplitString(mdsAddr, ",", &mdsAddrVec_);

if (mdsAddrVec_.empty()) {

std::cout << "Split mds address fail!" << std::endl;

return -1;

}

int res = InitDummyServerMap(dummyPort);

if (res != 0) {

std::cout << "init dummy server map fail!" << std::endl;

return -1;

}

for (uint64_t i = 0; i < mdsAddrVec_.size(); ++i) {

if (channel_.Init(mdsAddrVec_[i].c_str(), nullptr) != 0) {

std::cout << "Init channel to " << mdsAddr << "fail!" << std::endl;

continue;

}

// 寻找哪个mds存活

curve::mds::topology::ListPhysicalPoolRequest request;

curve::mds::topology::ListPhysicalPoolResponse response;

curve::mds::topology::TopologyService_Stub stub(&channel_);

brpc::Controller cntl;

cntl.set_timeout_ms(FLAGS_rpcTimeout);

stub.ListPhysicalPool(&cntl, &request, &response, nullptr);

if (cntl.Failed()) {

continue;

}

currentMdsIndex_ = i;

isInited_ = true;

return 0;

}

std::cout << "Init channel to all mds fail!" << std::endl;

return -1;

}

in tools-v1(snapshot):

curve/src/tools/snapshot_clone_client.cpp

Lines 28 to 42 in 4e73df5

int SnapshotCloneClient::Init(const std::string& serverAddr,

const std::string& dummyPort) {

curve::common::SplitString(serverAddr, ",", &serverAddrVec_);

if (serverAddrVec_.empty()) {

// no snapshot clone server

return 1;

}

int res = InitDummyServerMap(dummyPort);

if (res != 0) {

std::cout << "init dummy server map fail!" << std::endl;

return -1;

}

return 0;

}

in tools-v2(mds):

curve/tools-v2/pkg/config/bs.go

Lines 195 to 211 in 4e73df5

func AddBsUint64RequiredFlag(cmd *cobra.Command, name string, usage string) {

cmd.Flags().Uint64(name, uint64(0), usage+color.Red.Sprint("[required]"))

cmd.MarkFlagRequired(name)

err := viper.BindPFlag(BSFLAG2VIPER[name], cmd.Flags().Lookup(name))

if err != nil {

cobra.CheckErr(err)

}

}

// add flag option

// bs mds[option]

func AddBsMdsFlagOption(cmd *cobra.Command) {

AddBsStringOptionFlag(cmd, CURVEBS_MDSADDR, "mds address, should be like 127.0.0.1:6700,127.0.0.1:6701,127.0.0.1:6702")

}

func AddBsMdsDummyFlagOption(cmd *cobra.Command) {

AddBsStringOptionFlag(cmd, CURVEBS_MDSDUMMYADDR, "mds dummy address, should be like 127.0.0.1:6700,127.0.0.1:6701,127.0.0.1:6702")

}

I guess that the strings is from user input or config file like template.yaml:

curve/tools-v2/pkg/config/template.yaml

Lines 21 to 23 in 4e73df5

mdsAddr: 127.0.0.1:6700,127.0.0.1:6701,127.0.0.1:6702

mdsDummyAddr: 127.0.0.1:7700,127.0.0.1:7701,127.0.0.1:7702

etcdAddr: 127.0.0.1:23790,127.0.0.1:23791, 127.0.0.1:23792

How can i get the addr and dummy in snapshot tools-v2 implement?(add two kv pairs about snapshot in template.yaml?)

Yes, and you can refer to etcd or mds to obtain the relevant configuration process, which can be read from the configuration file or specified by flag.

Cyber-SiKu added the enhancement improve feature label Mar 30, 2023

Cyber-SiKu mentioned this issue Mar 30, 2023

[2023 Q2 Developer Activities]: Call For Participation！ #2334

Closed

wuhongsong added the good first issue Good for newcomers label Apr 10, 2023

Cyber-SiKu self-assigned this Apr 20, 2023

ilixiaocui assigned Xinlong-Chen and unassigned Cyber-SiKu Apr 23, 2023

This was referenced Apr 26, 2023

[feat] tools-v2: add snapshot-clone-status #2441

Closed

[feat] tools-v2: add snapshot-clone-status #2442

Merged

Cyber-SiKu closed this as completed May 8, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[curve/toos-v2]: add snapshot-clone-status #2346

[curve/toos-v2]: add snapshot-clone-status #2346

Cyber-SiKu commented Mar 30, 2023

zhanghuidinah commented Apr 12, 2023

Uh oh!

Xinlong-Chen commented Apr 21, 2023 •

edited

Loading

Uh oh!

ilixiaocui commented Apr 23, 2023

Uh oh!

Xinlong-Chen commented Apr 25, 2023 •

edited

Loading

Uh oh!

Cyber-SiKu commented Apr 25, 2023 •

edited

Loading

Uh oh!

Xinlong-Chen commented Apr 25, 2023 •

edited

Loading

Uh oh!

Cyber-SiKu commented Apr 25, 2023

Uh oh!

Xinlong-Chen commented Apr 25, 2023

Uh oh!

Cyber-SiKu commented Apr 25, 2023

Uh oh!

Cyber-SiKu commented Apr 25, 2023

Uh oh!

Xinlong-Chen commented Apr 25, 2023

Uh oh!

Cyber-SiKu commented Apr 26, 2023

Uh oh!

[curve/toos-v2]: add snapshot-clone-status #2346

[curve/toos-v2]: add snapshot-clone-status #2346

Comments

Cyber-SiKu commented Mar 30, 2023

zhanghuidinah commented Apr 12, 2023

Uh oh!

Xinlong-Chen commented Apr 21, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ilixiaocui commented Apr 23, 2023

Uh oh!

Xinlong-Chen commented Apr 25, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Cyber-SiKu commented Apr 25, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Xinlong-Chen commented Apr 25, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Cyber-SiKu commented Apr 25, 2023

Uh oh!

Xinlong-Chen commented Apr 25, 2023

Uh oh!

Cyber-SiKu commented Apr 25, 2023

Uh oh!

Cyber-SiKu commented Apr 25, 2023

Uh oh!

Xinlong-Chen commented Apr 25, 2023

Uh oh!

Cyber-SiKu commented Apr 26, 2023

Uh oh!

Xinlong-Chen commented Apr 21, 2023 •

edited

Loading

Xinlong-Chen commented Apr 25, 2023 •

edited

Loading

Cyber-SiKu commented Apr 25, 2023 •

edited

Loading

Xinlong-Chen commented Apr 25, 2023 •

edited

Loading