github.com/silveraid/fabric-ca@v1.1.0-preview.0.20180127000700-71974f53ab08/README.md (about) 1 # Fabric CA Developer's Guide 2 3 This is the Developer's Guide for Fabric CA, which is a Certificate Authority for Hyperledger Fabric. 4 5 See [User's Guide for Fabric CA](https://hyperledger-fabric-ca.readthedocs.io) for information on how to use Fabric CA. 6 7 ## Prerequisites 8 9 * Go 1.9+ installation or later 10 * **GOPATH** environment variable is set correctly 11 * docker version 17.03 or later 12 * docker-compose version 1.11 or later 13 * A Linux Foundation ID (see [create a Linux Foundation ID](https://identity.linuxfoundation.org/)) 14 15 16 ## Contribution guidelines 17 18 You are welcome to contribute to Fabric CA! 19 20 The following are guidelines to follow when contributing: 21 22 1. See the general information about [contributing to fabric](http://hyperledger-fabric.readthedocs.io/en/latest/CONTRIBUTING.html). 23 24 2. To set up your development environment for doing common development tasks, see [bash_profile](https://github.com/hyperledger/fabric-ca/blob/master/scripts/bash_profile). This contains variables and functions which can be copied directly into your `.bash_profile` file. Even if you do not use bash, you should still find the functions instructive. For example: 25 a. **clone** - pulls the latest fabric-ca code from gerrit and places it based on your GOPATH setting 26 b. **cdr** - cd to the fabric-ca repository root, which is equivalent to "cd $GOPATH/src/github.com/hyperledger/fabric-ca" 27 c. **gencov** - generates a test coverage report 28 29 3. To run the unit tests manually: 30 31 ``` 32 # cdr 33 # make unit-tests 34 ``` 35 36 The test coverage for each package must be 75% or greater. If this fails due to insufficient test coverage, then you can run `gencov` to get a coverage report to see what code is not being tested. Once you have added additional test cases, you can run `go test -cover` in the appropriate package to see the current coverage level. 37 38 WARNING: Running the unit-tests may fail due to too many open file descriptors. 39 Depending on where the failure occurs, the error message may not be obvious and may only say something similar to "unable to open database file". 40 Depending on the settings on your host, you may need to increase the maximum number of open file descriptors. 41 For example, the OSX default per-process maximum number of open file descriptors is 256. 42 You may issue the following command to display your current setting: 43 44 ``` 45 # ulimit -n 46 256 47 ``` 48 49 And the following command will increase this setting to 65536: 50 51 ``` 52 # ulimit -n 65536 53 ``` 54 55 Please note that this change is only temporary. To make it permanent, you will need to consult the documentation for your host operating system. 56 57 ## Package overview 58 59 1. **cmd/fabric-ca-server** contains the main for the fabric-ca-server command. 60 2. **cmd/fabric-ca-client** contains the main for the fabric-ca-client command. 61 3. **lib** contains most of the code. 62 a) **server.go** contains the main Server object, which is configured by **serverconfig.go**. 63 b) **client.go** contains the main Client object, which is configured by **clientconfig.go**. 64 4. **lib/csp** contains some functions related to the Crypto Service Provider. 65 5. **lib/dbutil** contains database utility functions. 66 6. **lib/ldap** contains LDAP client code. 67 7. **lib/spi** contains Service Provider Interface code for the user registry. 68 8. **lib/tls** contains TLS related code for server and client. 69 9. **util** contains various utility functions. 70 71 ## Additional info 72 73 ## Profiling 74 Fabric CA server can be profiled two ways, namely, using benchmarks and by retrieving profiling data from the server (at /debug/pprof/ endpoint) while running load. 75 76 ### Benchmarks 77 You can profile the benchmarks by running `make bench-cpu` or `make bench-mem` commands. You can profile benchmarks in one package or all the packages using these make targets. For example, to profile benchmarks in the *lib* package, run: `make bench-cpu pkg=github.com/hyperledger/fabric-ca/lib`. This will create **bench-cpu.prof**, **lib.test** and **bench** files in the *lib* folder. The **bench** file will contain benchmark stats: bytes/operation, allocations/operation and nanoseconds/operation. **lib.test** file is the executable and **bench-cpu.prof** contains cpu profile information. To analyze the profile, run: `go tool pprof lib.test bench-cpu.prof`. Similarly, you can run `make bench-mem pkg=github.com/hyperledger/fabric-ca/lib` to perform memory profiling of the benchmarks in the *lib* package. The **bench-mem.prof** file contains memory profile information. 78 79 If you run `make bench-cpu` or `make bench-mem` without *pkg* variable, benchmarks in each package are run with cpu or memory profiling. So, executable, benchmark output, and profile info files are created in each package folder. You need to analyze these profiles separately. 80 81 ### Whole server 82 To enable profiling on the server, set the FABRIC_CA_SERVER_PROFILE_PORT environment 83 variable to a valid, available port number and start the server. The server will start listening for profile requests at the */debug/pprof/* HTTP endpoint and the specified port. Then run `go tool pprof` with server's profiling URL (http://<server host>:<profiling port>/debug/pprof/<profile|heap|block>) as an argument, it will download and examine a live profile. 84 85 You can start the server in the FVT image by running following docker command from the fabric-ca root directory: 86 87 `docker run -p 8888:8888 -p 8054:8054 -v $PWD:/opt/gopath/src/github.com/hyperledger/fabric-ca -e FABRIC_CA_SERVER_PROFILE_PORT=8054 --name loadTest -td hyperledger/fabric-ca-fvt test/fabric-ca-load-tester/launchServer.sh 1` 88 89 Then start the load by running `/test/fabric-ca-load-tester/runLoad.sh -B` 90 91 In other window, you can start profiling by running (assuming load test takes about a minute to complete): 92 93 `curl http://localhost:8054/debug/pprof/profile?seconds=60 > load-cpu.prof` 94 95 then analyze the profile: 96 97 `go tool pprof bin/fabric-ca-server load-cpu.prof` 98 99 OR simply run: 100 101 `go tool pprof -seconds=60 -output=load-cpu.prof http://localhost:8054/debug/pprof/profile` 102 103 You can use commands like *top*, *top -cum*, *list* and *web* to look at the top consumers, list the code to see the hotspots and to view the graph in a browser. You can run `go tool pprof -h` to view all the options supported by the pprof tool 104 105 You can also use [**go-torch**](https://github.com/uber/go-torch) tool to analyze the profile: 106 107 `go-torch bin/fabric-ca-server load-cpu.prof` 108 109 ### Profiling Fabric CA client 110 To enable profiling on the client, set the FABRIC_CA_CLIENT_PROFILE_MODE environment variable to either "heap" or "cpu" to enable heap, cpu profiling respectively. A file containing profiling data is created in the present working directory of the client. Heap profiling data is written to **mem.pprof** and cpu profiling data is written to **cpu.pprof**. You can run `go tool pprof <client executable> <profiling file>` to analyze the profiling data. 111 112 ### Profiling links 113 https://blog.golang.org/profiling-go-programs 114 https://medium.com/@hackintoshrao/daily-code-optimization-using-benchmarks-and-profiling-in-golang-gophercon-india-2016-talk-874c8b4dc3c5 115 https://www.youtube.com/watch?v=2h_NFBFrciI 116 https://software.intel.com/en-us/blogs/2014/05/10/debugging-performance-issues-in-go-programs 117 http://www.soroushjp.com/2015/01/27/beautifully-simple-benchmarking-with-go/ 118 https://vinceyuan.github.io/profiling-memory-usage-of-a-go-app/ 119 https://www.youtube.com/watch?v=N3PWzBeLX2M&feature=youtu.be 120 https://www.youtube.com/watch?v=oorX84tBMqo&feature=youtu.be 121 122 ### FVT 123 124 See [FVT tests](scripts/fvt/README.md) for information on functional verification test cases. 125 126 127 ### Updating the cfssl vendored package 128 Following are the steps to update cfssl package using version 1.0.8 of govendor tool. 129 130 * Remove cfssl from vendor folder 131 * cd $GOPATH/src/github.com/hyperledger/fabric-ca/vendor 132 * govendor remove github.com/cloudflare/cfssl/... 133 * rm -rf github.com/cloudflare/cfssl/ 134 135 * Clone cfssl repo 136 * cd $GOPATH/src/github.com/ 137 * mkdir cloudflare 138 * cd cloudflare 139 * git clone https://github.com/cloudflare/cfssl.git 140 141 * Add cfssl from $GOPATH to the vendor folder 142 * cd $GOPATH/src/github.com/hyperledger/fabric-ca/vendor 143 * govendor add github.com/cloudflare/cfssl/^ 144 * You can optionally specify revision or tag to add a particular revision of code to the vendor folder 145 * govendor add github.com/cloudflare/cfssl/^@abc12032 146 147 * Remove sqlx package from cfssl vendor folder. This is because certsql.NewAccessor (called by fabric-ca) requires sqlx.db object to be passed from the same package. If we were to have sqlx package both in fabric-ca and cfssl vendor folder, go compiler will throw an error 148 * rm -rf github.com/cloudflare/cfssl/vendor/github.com/jmoiron/sqlx 149 150 * Remove the packages that are added to the fabric-ca vendor folder that are not needed by fabric-ca 151 152 153 ## License <a name="license"></a> 154 155 Hyperledger Project source code files are made available under the Apache License, Version 2.0 (Apache-2.0), located in the [LICENSE](LICENSE) file. Hyperledger Project documentation files are made available under the Creative Commons Attribution 4.0 International License (CC-BY-4.0), available at http://creativecommons.org/licenses/by/4.0/.