5 - Better error message if daemon/scripts-security-upd
6 is not on scripts-security-upd list
8 - Consider adding user config to repository
9 - Put php.ini links in repos (probably will need another
11 - Version configuration
13 - Consider making usermode wizard operation a support mode
14 (mostly for letting users upgrade things themself)
16 - Add repository flag to migrate so that we can specify an
17 arbitrary repository to migrate to
19 - The great initial deploy:
20 - Turn on mediawiki new autoinstaller
21 - Migrate all mediawik installs
23 - Implement proper deploy log parsing; this basically means we
24 need to be able to introspect Git Log. Consider using git-python
25 for this. There's also missing functionality, bad error handling
26 and hacks in the prototype implementation of upgrade
28 - Wordpress needs to have .scripts dir in all -scripts versions
29 (remember --no-walk!) (also make .scripts/.htaccess)
30 - Wordpress needs to have a .scripts/update script written for
31 its latest version (do this after its migration)
32 - Wordpress needs to check for php.ini files (which it almost
33 certianly has) and commit messages
35 - Summary script should be more machine friendly, and should not
36 output summary charts when I increase specificity
37 - Check how many autoinstalls are missing w bits for
38 daemon.scripts (this would need pyafs)
39 - Consider fixing Wizard's commit messages
41 PULLING OUT CONFIGURATION FILES IN AN AUTOMATED MANNER
43 advancedpoll: Template file to fill out
44 django: Noodles of template files
45 gallery2: Multistage install process
47 mediawiki: One-step install process
48 phpbb: Multistage install process
49 phpical: Template file
52 wordpress: Multistage install process
54 PHILOSOPHY ABOUT LOGGING
56 Logging is most useful when performing a mass run. This
57 includes things such as mass-migration as well as when running
58 summary reports. An interesting property about mass-migration
59 or mass-upgrade, however, is that if they fail, they are
60 idempotent, so an individual case can be debugged simply running
61 the single-install equivalent with --debug on. (This, indeed,
62 may be easier to do than sifting through a logfile).
64 It is a different story when you are running a summary report:
65 you are primarily bound by your AFS cache and how quickly you can
66 iterate through all of the autoinstalls. Checking if a file
67 exists on a cold AFS cache may
68 take several minutes to perform; on a hot cache the same report
69 may take a mere 3 seconds. When you get to more computationally
70 expensive calculations, however, even having a hot AFS cache
71 is not enough to cut down your runtime.
73 There are certain calculations that someone may want to be
74 able to perform on manipulated data. As such, this data should
75 be cached on disk, if the process for extracting this data takes
76 a long time. Also, for usability sake, Wizard should generate
77 the common case reports.
79 Ensuring that machine parseable reports are made, and then making
80 the machinery to reframe this data, increases complexity. Therefore,
81 the recommendation is to assume that if you need to run iteratively,
82 you'll have a hot AFS cache at your fingerprints, and if that's not
83 fast enough, then cache the data.
85 COMMIT MESSAGE FIELDS:
87 Installed-by: username@hostname
88 Pre-commit-by: Real Name <username@mit.edu>
89 Upgraded-by: Real Name <username@mit.edu>
90 Migrated-by: Real Name <username@mit.edu>
91 Wizard-revision: abcdef1234567890
92 Wizard-args: /wizard/bin/wizard foo bar baz
96 Committer: Real Name <username@mit.edu>
97 Author: lockername locker <lockername@scripts.mit.edu>
101 - A perfectly formed autoinstall with upgrade paths for all of
102 the intervening versions is not really feasible to implement.
103 As such, we want to migrate everything to -scripts, and then
104 generate a -scripts2 with the correct .scripts directory.
105 We will then nop update some installs, but this will prevent
106 us from having to migrate and update concurrently. Treat
107 a scripts2 upgrade from migration the same way you would treat
108 a botched scripts upgrade.
110 - Currently all repositories are initialized with --shared, which
111 means they have basically ~no space footprint. However, it
112 also means that /mit/scripts/wizard/srv MUST NOT lose revs.
114 - Full fledged logging options. Namely:
115 x all loggers (delay implementing this until we actually have debug stmts)
117 - debug => loglevel = DEBUG
119 - default is WARNING (see below for exception)
120 - verbose => loglevel = INFO
121 x file logger (only allowed for serial processing)
123 - log-file => loglevel = INFO
124 x database logger (necessary for parallel processing, not implemented)
126 - log-db => loglevel = INFO
128 - More on the database logger: it will be very simple with one
129 table named `logs` in SQLite, with columns: `job`, `level`,
130 `message`. Job identifies the subprocess/thread that emitted
131 the log, so things can be correlated together. We will then
132 have `wizard dump` which takes a database like this and dumps
133 it into a file logger type file. The database may also store
134 a queue like structure which can be used to coordinate jobs.
138 * Some parts of the infrastructure will not be touched, although I plan
139 on documenting them. Specifically, we will be keeping:
141 - parallel-find.pl, and the resulting
142 /mitalso make .scripts/.htaccess/scripts/sec-tools/store/scriptslist
143 This script might need to be adapted if we decide to nuke
144 .scripts-version files.
146 - The current install scripts will be kept in place, sans changes
147 necessary to make them use Git install of copying the script over.
148 Porting these scripts to Python and making them modular would be
149 nice, but is priority. For the long term, seeing this scripts
150 be packaged with rest of our code would be optimal.
152 * The new procedure for generating an update is as follows:
153 (check out the mass-migration instructions for something in this spirit,
154 although uglier in some ways)
156 0. ssh into not-backward, temporarily give the daemon.scripts-security-upd
157 bits by blanching it on system:scripts-security-upd, and run parallel-find.pl
159 1. Have the Git repository and working copy for the project on hand.
161 2. Checkout the pristine branch
163 3. Remove all files from the working copy. Use `wipe-working-dir`
165 4. Download the new tarball
167 5. Extract the tarball over the working copy (`cp -R a/. b` works well,
168 remember that the working copy is empty)
170 6. Check for empty directories and add stub files as necessary.
171 Use `preserve-empty-dir`
173 7. Git add it all, and then commit as a new pristine version (v1.2.3)
175 8. Checkout the master branch
177 9. [FOR EXISTING REPOSITORIES]
178 Merge the pristine branch in. Resolve any conflicts that our
179 patches have with new changes. Do NOT let Git auto-commit it
180 with --no-commit (otherwise, you want to git commit --amend
181 to keep our history clean
183 [FOR NEW REPOSITORIES]
184 See if any patches are needed to make this run smoothly on
187 [FOR NEW REPOSITORIES]
189 echo "Deny from all" > .scripts/.htaccess
190 touch .scripts/update
191 chmod a+x .scripts/update
193 10. Check if there are any special update procedures, and update/create the
194 .scripts/update shell script as necessary (this means that any
195 application specific update logic will be kept with the actual
196 source code. The language of this update script will vary
197 depending on context.)
199 11. Commit your changes, and tag as v1.2.3-scripts (or scripts2, if
200 you are amending an install without an upstream changes)
202 12. Test the new update procedure using
203 `wizard upgrade --with=/path/to/repo /your/autoinstall` (this will
204 read out master as your "latest" version).
205 Use git commit --amend to fix any bugs (alternatively, squash them
208 13. You can also do a "mass" version of this using:
209 `wizard -d testbed.txt massupgrade --with=/path/to/repo app`
210 You'll need perms for any testbed stuff you want.
212 GET APPROVAL BEFORE PROCEEDING ANY FURTHER
214 NOTE: The following commands are to be run on not-backward.mit.edu.
215 You'll need to add daemon.scripts-security-upd to
216 scripts-security-upd to get bits to do this. Make sure you remove
217 these bits when you're done.
219 14. Run `wizard research appname`
220 which uses Git commands to check how many
221 working copies apply the change cleanly, and writes out a logfile
222 with the working copies that don't apply cleanly. It also tells
223 us about "corrupt" working copies.
225 15. Run `wizard massupgrade appname`, which applies the update to all working
226 copies possible, and sends mail to users to whom the working copy
227 did not apply cleanly. It also frobs .scripts-version for successful
228 upgrades (maybe not, depending on our plans).
230 16. Run parallel-find.pl to update our inventory
232 * For mass importing into the repository, the steps are:
233 (this probably won't ever be automated, becuase there are fiddly bits)
236 # let app-1.2.3 be the scripts folder originally in deploydev
237 # let this folder be srv/
238 # you can also do a git clone
243 unfurl app-1.2.3 app # [FIDDLY BIT]
244 # NOTE: contents of application are now in app directory
247 git commit -s -m "App 1.2.3"
250 # NOTE: you're still on master branch
251 # WARNING: the following operation might require -p1
252 patch -p0 < ../app-1.2.3/app-1.2.3.patch # [FIDDLY BIT]
253 # NOTE: please sanity check the patch!
255 # NOTE: -a flag is to handle if the patch deleted something
256 git commit -as -m "App 1.2.3-scripts"
257 git tag v1.2.3-scripts
260 # let this folder be srv/app.git
261 git checkout pristine
262 # NOTE: this preserves your .git folder, but removes everything
265 unfurl app-1.2.3 app # [FIDDLY BIT]
267 # NOTE: please sanity check app directory
269 # NOTE: -a is to take care of deletions
270 git commit -as -m "App 1.2.3"
272 [FIDDLE AROUND. FIDDLE AROUND]
273 [IF THE PATCH HAS CHANGED]
274 # You are on the pristine branch
275 # NOTE: Now, the tricky part (this is different from a real update)
276 git symbolic-ref HEAD refs/heads/master
277 # NOTE: Now, we think we're on the master branch, but we have
278 # pristine copy checked out
279 # NOTE: -p0 might need to be twiddled
280 patch -p0 < ../app-1.2.3/app-1.2.3.patch
282 # COMMENT: used to git checkout .scripts here
283 # then check if the directory needs an updated update script
284 # NOTE: Fake the merge
285 git rev-parse pristine > .git/MERGE_HEAD
286 [IF THE PATCH HASN'T CHANGED]
288 git merge --no-commit pristine
289 git commit -as -m "App 1.2.3-scripts"
290 git tag v1.2.3-scripts
292 * The repository for a given application will contain the following files:
294 - The actual application's files, as from the official tarball
296 - A .scripts directory, which contains the following information:
298 * .scripts/update shell script (with the +x bit set appropriately),
299 which performs the commands necessary to update a script. This can
302 * .scripts/.htaccess to prevent this directory from being accessed
305 * .scripts/database (generated) contains the database the
306 user installed the script to, so scripts-remove can clean it
308 XXX: Could cause problems if a user copies the autoinstall,
309 fiddles with the DB credentials, and then scripts-remove's
310 the autoinstall. Possible fix is to add the original
311 directory as a sanity check. Additionally, we could have
312 the application read out of this file.
314 * .scripts/old-version (optional) the old value of .scripts-versoin
316 * .scripts/install (eventually) interactively installs the
317 applicatoin from command line.
319 * The autoupgrade shall be the process of:
321 # Make the directory not accessible by the outside world (htaccess, but be careful!)
323 git commit -m 'automatically generated backup'
324 git pull origin master
325 if [ $? ne 0 ]; then git reset --hard; echo 'conflicts during upgrade'; fi
329 (with some more robust error checking, a proper dry run mechanism to, and
332 * Make 'wizard summary' generate nice pretty graphs of installs by date
333 (more histograms, will need to check actual .scripts-version files.)
335 * Update AFS patch to advertise its existence, so we can check for it