Browse Source

update gin-cli troubleshooting

Keisuke Sehara 3 years ago
parent
commit
8bf25dcb56
4 changed files with 182 additions and 2 deletions
  1. 2 1
      README.md
  2. 1 1
      gin-client/basics.md
  3. 3 0
      gin-client/tips.md
  4. 176 0
      gin-client/troubleshooting.md

+ 2 - 1
README.md

@@ -18,7 +18,8 @@ introduction to how to deal with gin (web interface, command-line, and probably
     - [Where to find basic information](gin-client/basics.md)
     - [Setting up for the wired network in Charité](gin-client/Charite-network.md)
     - (Advanced) Dealing with [what is annexed, and what is not](https://git-annex.branchable.com/tips/largefiles/)
-    - [**Tips and troubleshooting**](gin-client/tips.md)
+    - [**Tips**](gin-client/tips.md)
+    - [**troubleshooting**](gin-client/troubleshooting.md)
 4. **WinGIN (GUI on Windows)**
     - What WinGIN does
     - Download and setup

+ 1 - 1
gin-client/basics.md

@@ -8,5 +8,5 @@ There is a [wiki page](https://gin.g-node.org/G-Node/Info/wiki/) that covers som
 - [How to commit changes to the local repository](https://gin.g-node.org/G-Node/Info/wiki/GIN+CLI+Usage+Tutorial#the-workflow)
 - [Synchronizing changes with the server (upload, download)](https://gin.g-node.org/G-Node/Info/wiki/GIN+CLI+Usage+Tutorial#the-workflow)
 
-You can refer to our [tips and troubleshooting](tips.md) page, too!
+You can refer to our [tips](tips.md) and [troubleshooting](troubleshooting.md) pages, too!
 

+ 3 - 0
gin-client/tips.md

@@ -0,0 +1,3 @@
+# Gin-CLI: Tips
+
+(TODO)

+ 176 - 0
gin-client/troubleshooting.md

@@ -0,0 +1,176 @@
+# Gin-CLI: troubleshooting
+
+- [The gin command not found](#the-gin-command-not-found)
+- [Gin-CLI hangs when adding / committing changes](#gin-cli-hangs-when-adding-committing-changes)
+- [Uploading fails with an error](#uploading-fails-with-an-error)
+- [The content of a large file is not uploaded](#the-content-of-a-large-file-is-not-uploaded)
+
+
+
+## The `gin` command not found
+
+### Symptoms
+
+Your terminal emulator complains that `gin` command is not found.
+
+But you can see your `gin` program on your PC.
+
+### Diagnosis
+
+In short: it is very likely that the `PATH` environment variable is not set (properly) for your Gin-CLI.
+
+The terminal emulator looks for commands by visiting the directories listed in its settings. The settings for the terminal emulator is called the "environment variables", and the variable for the list of the command-lookup directories is called `PATH`.
+
+In many cases, you have to manually add to `PATH` the directory where the command is located.
+
+### Solutions
+
+- If there has been some time since you downloaded Gin-CLI, **downloading the latest binary** will prevent any further unexpected problems.
+- For Gin-CLI on Windows, there is a nice file `set-global.bat` to do it for you. Just double-click this file.
+
+Even when you use `set-global.bat`, sometimes the directory is not added properly because there are already lots of directories listed in `PATH` and it cannot add any more due to the character-number limitation(!). It can occur when you have already installed lots of programs on your computer.
+
+In such cases, the possibilities are: 
+
+- Check `PATH` one by one, and try to reduce the number of characters in it:
+  defining the `PATH` environment variable is much like defining the `HTTP_PROXY` variable on [this section](Charite-network.md#setting-up-the-http-proxy), so please refer to it.
+- Use `gin-shell.bat` as an alternative terminal emulator.
+
+
+
+## Gin-CLI hangs when adding / committing changes
+
+### Symptoms
+
+- You run either `gin commit` or `gin upload`  to add/commit changes.
+- Gin-CLI lists up the files that changed, but the progress stops at some point onwards.
+- You can abort the procedure by hitting `Ctrl + C`, but you end up failing to add changes.
+
+### Diagnosis
+
+Although **this situation has not been fully resolved**, there are two factors that may be involved: large file-sizes, and a large number of files.
+
+- **large file size**: adding a large file via `gin` takes time (increases linearly with the file size). Probably you see in the Process Manager that Gin-CLI consumes some CPU power until it is done. Please wait for some time (up to several minutes).
+- **many files**: 
+  - I suspect that `gin` has some issues when adding too many changes all at once.
+  - Git is slow when you commit a lot of changes all at once (takes up to several minutes, hopefully).
+  - Git for Windows cannot handle a commit with more than 2 GB (excluding the sizes of "large files").
+
+### Solutions (or tips)
+
+Below are some tips to avoid the situation:
+
+- **add, commit and upload separately**.
+  1. to add changes: `gin annex add <files or directories>`
+  2. to add deletions: `gin git add --update`
+  3. to commit changes: `gin git commit` (optionally appended by a custom message: `-m "<message>"`)
+  4. to upload: `gin upload` (note: without a dot `.` at the end)
+- **try to add / commit large directories separately**: individual processes will take (at least) much shorter time, and thereby keeping you feel safer and more comfortable.
+- **register all "data files" as "large" (annexed) files**: to do so, under the repository-root directory (or optionally under the individual directories separately) where you keep data files, add and edit the `.gitattributes` file.
+  **IMPORTANT**: if you want `gin` to take this settings into account, you _must_ use `gin annex add` (instead of `gin commit`).
+  1. Create a plain-text file, and rename it to `.gitattributes` (_without any extension_ e.g. ".txt").
+  2. Open it from your text editor, and add lines as below:
+     - The setting in general: a line `* annex.largefiles=largerthan=100kb` will add all files into annex if they have more than 100 kilobytes in size. 
+     - The extension-specific setting: for example, a line `*.mat annex.largefiles=anything` will add all files into annex if they have the ".mat" extension.
+     - The file-specific un-setting: for example, a line `*.ipynb annex.largefiles=nothing` will add the ".ipynb" file normally (i.e. without being annexed) no matter the file size.
+     - You can add a comment line, by starting it with a hash `#`.
+
+The example file is as below (you can copy and paste it if you like):
+
+```shell
+# general settings
+*       annex.largefiles=largerthan=100kb
+LICENSE annex.largefiles=nothing
+
+# data files
+*.dat   annex.largefiles=anything
+*.mat   annex.largefiles=anything
+*.ibw   annex.largefiles=anything
+*.ifn   annex.largefiles=anything
+*.npy   annex.largefiles=anything
+*.npz   annex.largefiles=anything
+*.tif   annex.largefiles=anything
+*.tiff  annex.largefiles=anything
+*.png   annex.largefiles=anything # it depends, but safer this way
+*.jpg   annex.largefiles=anything # it depends, but safer this way
+*.jpeg  annex.largefiles=anything # it depends, but safer this way
+*.bmp   annex.largefiles=anything
+*.avi   annex.largefiles=anything
+*.mp4   annex.largefiles=anything
+*.json  annex.largefiles=anything # it depends, but safer this way
+*.csv   annex.largefiles=anything # it depends, but safer this way
+*.tsv   annex.largefiles=anything # it depends, but safer this way
+*.xml   annex.largefiles=anything # it depends, but safer this way
+
+# binary document files
+*.pdf   annex.largefiles=anything
+*.docx  annex.largefiles=anything
+*.pptx  annex.largefiles=anything
+*.xlsx  annex.largefiles=anything
+
+# code and text files
+*.md    annex.largefiles=nothing
+*.txt   annex.largefiles=nothing
+*.yml   annex.largefiles=nothing
+*.yaml  annex.largefiles=nothing
+*.m     annex.largefiles=nothing
+*.r     annex.largefiles=nothing
+*.py    annex.largefiles=nothing
+*.ipynb annex.largefiles=nothing
+*.htm   annex.largefiles=nothing
+*.html  annex.largefiles=nothing
+```
+
+
+
+## Uploading fails with an error
+
+Uploading involves a lot of uncertainties, but `gin` may only say `upload failed` (if it says `upload OK`, then please refer to [another section](#the-content-of-a-large-file-is-not-uploaded)).
+
+Please check the points below, one by one:
+
+1. **Is the Gin server alive?**
+   - Visit https://gin.g-node.org/ and check if there are any maintenance etc. going on.
+2. **Have you changed your repository name recently?**
+   It changes the URL to communicate, so the local repositories that were downloaded before renaming will fail to upload/download changes.
+   - In your local repository: run `gin git remote set-url origin <URL>`
+     The `<URL>` must be in the form: `ssh://git@gin.g-node.org:22/<organization>/<repository-name>` . **Make sure that you do not have `.git` after the repository name!**
+3. **Do you use the wired connection from Charité?**
+   - Please refer to [this page](Charite-network.md) for problems related to this point.
+   - In case you have not done any proxy settings on the command line, try connecting via a Wi-Fi.
+   - If the error says there is a problem in connecting to `larkumlab-data-micro.charite.de`, then it is the problem with the SSH proxy server. Ask someone for restarting the proxy service.
+   - Please do note that some other problems (like [this](#the-content-of-a-large-file-is-not-uploaded)) may persist even after solving this problem...
+
+If the problem does not fall into one of the points above, please contact someone who knows more.
+
+
+
+## The content of a large file is not uploaded
+
+### Symptoms
+
+- You run `gin upload`, and it says `uploading OK`.
+- The directory organization becomes up-to-date on the server after `gin upload`.
+- Yet, if you look at the large file on the server, it says "no content available".
+
+### Diagnosis
+
+Below are the points to our current interpretation:
+
+- `gin` has experienced a problem (likely with SSH) some time before, when it tries to upload contents of large data files.
+- and then `gin` erranously considered "this server is not compatible with uploading large data files".
+- since then, `gin` does not attempt to upload file contents anymore.
+
+### Solution
+
+In short: `gin` (or git annex) uses the git config `remote.origin.annex-ignore` to remember whether the server has the annex backend. All you have to do is to unset this config.
+
+In reality, it requires a bit of complexity to solve the problem:
+
+1. In your repository, run `gin git config --list`. It shows all the repository-specific Git configuration.
+   (listed in the alphabetical order)
+2. Find the line starting with `remote.origin.annex-ignore=`
+   Make sure that is says `true` (Otherwise it must be due to some other reasons).
+3. Run: `gin git config --unset remote.origin.annex-ignore`
+   It should return without any output.
+4. You can run `gin git config --list` to make sure that the property is gone.