Summary:
The search indexing API has several problems right now:
- Always runs in-process.
- It would be nice to push this into the task queue for performance. However, the API currently passses an object all the way through (and some indexers depend on preloaded object attributes), so it can't be dumped into the task queue at any stage since we can't serialize it.
- Being able to use the task queue will also make rebuilding indexes faster.
- Instead, make the API phid-oriented.
- No uniform indexing API.
- Each "Editor" currently calls SomeCustomIndexer::indexThing(). This won't work with AbstractTransactions. The API is also just weird.
- Instead, provide a uniform API.
- No uniform CLI.
- We have `scripts/search/reindex_everything.php`, but it doesn't actually index everything. Each new document type needs to be separately added to it, leading to stuff like D3839. Third-party applications can't provide indexers.
- Instead, let indexers expose documents for indexing.
- Not application-oriented.
- All the indexers live in search/ right now, which isn't the right organization in an application-orietned view of the world.
- Instead, move indexers to applications and load them with SymbolLoader.
Test Plan:
- `bin/search index`
- Indexed one revision, one task.
- Indexed `--type TASK`, `--type DREV`, etc., for all types.
- Indexed `--all`.
- Added the word "saboteur" to a revision, task, wiki page, and question and then searched for it.
- Creating users is a pain; searched for a user after indexing.
- Creating commits is a pain; searched for a commit after indexing.
- Mocks aren't currently loadable in the result view, so their indexing is moot.
Reviewers: btrahan, vrana
Reviewed By: btrahan
CC: 20after4, aran
Maniphest Tasks: T1991, T2104
Differential Revision: https://secure.phabricator.com/D4261
Summary:
- Build "sshd-auth" (for authentication) and "sshd-exec" (for command execution) binaries. These are callable by "sshd-vcs", located [[https://github.com/epriestley/sshd-vcs | in my account on GitHub]]. They are based on precursors [[https://github.com/epriestley/sshd-vcs-glue | here on GitHub]] which I deployed for TenXer about a year ago, so I have some confidence they at least basically work.
- The problem this solves is that normally every user would need an account on a machine to connect to it, and/or their public keys would all need to be listed in `~/.authorized_keys`. This is a big pain in most installs. Software like Gitosis/Gitolite solve this problem by giving you an easy way to add public keys to `~/.authorized_keys`, but this is pretty gross.
- Roughly, instead of looking in `~/.authorized_keys` when a user connects, the patched sshd instead runs `echo <public key> | sshd-auth`. The `sshd-auth` script looks up the public key and authorizes the matching user, if they exist. It also forces sshd to run `sshd-exec` instead of a normal shell.
- `sshd-exec` receives the authenticated user and any command which was passed to ssh (like `git receive-pack`) and can route them appropriately.
- Overall, this permits a single account to be set up on a server which all Phabricator users can connect to without any extra work, and which can safely execute commands and apply appropriate permissions, and disable users when they are disabled in Phabricator and all that stuff.
- Build out "sshd-exec" to do more thorough checks and setup, and delegate command execution to Workflows (they now exist, and did not when I originally built this stuff).
- Convert @btrahan's conduit API script into a workflow and slightly simplify it (ConduitCall did not exist at the time it was written).
The next steps here on the Repository side are to implement Workflows for Git, SVN and HG wire protocols. These will mostly just proxy the protocols, but also need to enforce permissions. So the approach will basically be:
- Implement workflows for stuff like `git receive-pack`.
- These workflows will implement enough of the underlying protocol to determine what resource the user is trying to access, and whether they want to read or write it.
- They'll then do a permissons check, and kick the user out if they don't have permission to do whatever they are trying to do.
- If the user does have permission, we just proxy the rest of the transaction.
Next steps on the Conduit side are more simple:
- Make ConduitClient understand "ssh://" URLs.
Test Plan: Ran `sshd-exec --phabricator-ssh-user epriestley conduit differential.query`, etc. This will get a more comprehensive test once I set up sshd-vcs.
Reviewers: btrahan, vrana
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T603, T550
Differential Revision: https://secure.phabricator.com/D4229
Summary: Add a bin/drydock symlink and break it into workflows. Nothing too special here.
Test Plan: Ran `bin/drydock wait-for-lease`, `bin/drydock lease`, `bin/drydock help`, etc.
Reviewers: btrahan
Reviewed By: btrahan
CC: aran
Maniphest Tasks: T2015
Differential Revision: https://secure.phabricator.com/D3867
Summary: Quora requested this (moving to S3) but it's also clearly a good idea.
Test Plan:
Ran with various valid/invalid options to test options. Error/sanity checking seemed OK.
Migrated individual local files.
Migrated all my local files back and forth between engines several times.
Uploaded some new files.
Reviewers: btrahan, vrana
Reviewed By: vrana
CC: aran
Maniphest Tasks: T1950
Differential Revision: https://secure.phabricator.com/D3808
Summary:
Basic "Fact" application with some storage, part of a daemon, and a control binary.
= Goals =
The general idea is that we have various statistics we'd like to compute, like the frequency of image macros, reviewer responsiveness, task close rates, etc. Computing these on page load is expensive and messy. By building an ETL pipeline and running it in a daemon, we can precompute statistics and just pull them out of "stats" tables.
One way to do this is just to completely hard-code everything, e.g. have a daemon that runs every hour which issues a big-ass query and dumps results into a table per-fact or per fact-group. But this has a bunch of drawbacks: adding new stuff to the pipeline is a pain, various fact aggregators can't share much code, updates are slow and expensive, we can never build generic graphs on top of it, etc.
I'm hoping to build an ETL pipeline which is generic enough that we can use it for most things we're interested in without needing schema changes, and so that installs can use it also without needing schema changes, while still being specific enough that it's fast and we can build useful stuff on top of it. I'm not sure if this will actually work, but it would be cool if it does so I'm starting pretty generally and we'll see how far I get. I haven't built this exact sort of thing before so I might be way off.
I'm basing the whole thing on analyzing entire objects, not analyzing changes to objects. So each part of the pipeline is handed an object and told "analyze this", not handed a change. It pretty much deletes all the old data about that thing and then writes new data. I think this is simpler to implement and understand, and it protects us from all sorts of weird issues where we end up with some kind of garbage in the DB and have to wipe the whole thing.
= Facts =
The general idea is that we extract "facts" out of objects, and then the various view interfaces just report those facts. This change has on type of fact, a "raw fact", which is directly derived from an object. These facts are concerete and relate specifically to the object they are derived from. Some examples of such facts might be:
D123 has 9 comments.
D123 uses macro "psyduck" 15 times.
D123 adds 35 lines.
D123 has 5 files.
D123 has 1 object.
D123 has 1 object of type "DREV".
D123 was created at epoch timestamp 89812351235.
D123 was accepted by @alincoln at epoch timestamp 8397981839.
The fact storage looks like this:
<factType, objectPHID, objectA, valueX, valueY, epoch>
Currently, we supprot one optional secondary key (like a user PHID or macro PHID), two optional integer values, and an optional timestamp. We might add more later. Each fact type can use these fields if it wants. Some facts use them, others don't. For instance, this diff adds a "N:*" fact, which is just the count of total objects in the system. These facts just look like:
<"N:*", "PHID-xxxx-yyyy", ...>
...where all other fields are ignored. But some of the more complex facts might look like:
<"DREV:accept", "PHID-DREV-xxxx", "PHID-USER-yyyy", ..., ..., nnnn> # User 'yyyy' accepted at epoch 'nnnn'.
<"FILE:macro", "PHID-DREV-xxxx", "PHID-MACR-yyyy", 17, ..., ...> # Object 'xxxx' uses macro 'yyyy' 17 times.
Facts have no uniqueness constraints. For @vrana's reviewer responsiveness stuff, we can insert multiple rows for each reviewer, e.g.
<"DREV:reviewed", "PHID-DREV-xxxx", "PHID-USER-yyyy", nnnn, ..., mmmm> # User 'yyyy' reviewed revision 'xxxx' after 'nnnn' seconds at 'mmmm'.
The second value (valueY) is mostly because we need it if we sample anything (valueX = observed value, valueY = sample rate) but there might be other uses. We might need to add "objectB" at some point too -- currently we can't represent a fact like "User X used macro Y on revision Z", so it would be impossible to compute macro use rates //for a specific user// based on this schema. I think we can start here though and see how far we get.
= Aggregated Facts =
These aren't implemented yet, but the idea is that we can then take the "raw facts" and compute derived/aggregated/rollup facts based on the raw fact table. For example, the "count" fact can be aggregated to arrive at a count of all objects in the system. This stuff will live in a separate table which does have uniqueness constraints, and come in the next diff.
We might need some kind of time series facts too, not sure about that. I think most of our use cases today are covered by raw facts + aggregated facts.
Test Plan: Ran `bin/fact` commands and verified they seemed to do reasonable things.
Reviewers: vrana, btrahan
Reviewed By: vrana
CC: aran, majak
Maniphest Tasks: T1562
Differential Revision: https://secure.phabricator.com/D3078
Summary:
Nothing new or exciting here yet, just moving the random scripts/repositories/ things to bin/repository. Also add `repository list`.
(Console stuff comes from D2841.)
Test Plan: Ran `repository list`, `repository pull`, `repository discover`, `repository discover --verbose`, `repository help`.
Reviewers: jungejason, vrana
Reviewed By: vrana
CC: aran
Differential Revision: https://secure.phabricator.com/D2849
Summary:
This addresses three issues with the current patch management system:
# Two people developing at the same time often pick the same SQL patch number, and then have to go rename it. The system catches this, but it's silly.
# Second/third-party developers can't use the same system to manage auxiliary storage they may want to add.
# There's no way to build mock databases for unit tests that need to do reads.
To resolve these things, you can now name your patches whatever you want and conflicts are just merge conflicts, which are less of a pain to fix than filename conflicts.
Dependencies are now a DAG, with implicit dependencies created on the prior patch if no dependencies are specified. Developers can add new concrete subclasses of `PhabricatorSQLPatchList` to add storage management, and define the dependency branchpoint of their patches so they apply in the correct order (although, generally, they should not depend on the mainline patches, presumably).
The commands `storage upgrade --namespace test1234` and `storage destroy --namespace test1234` will allow unit tests to build and destroy MySQL storage.
A "quickstart" mode allows an upgrade from scratch in ~1200ms. Destruction takes about 200ms. These seem like fairily reasonable costs to actually use in tests. Building from scratch patch-by-patch takes about 6000ms.
Test Plan:
- Created new databases from scratch with and without quickstart in a separate test namespace. Pointed the webapp at the test namespaces, browsed around, everything looked good.
- Compared quickstart and no-quickstart dump states, they're identical except for mysqldump timestamps and a few similar things.
- Upgraded a legacy database to the new storage format.
- Destroyed / dumped storage.
Reviewers: edward, vrana, btrahan, jungejason
Reviewed By: btrahan
CC: aran, nh
Maniphest Tasks: T140, T345
Differential Revision: https://secure.phabricator.com/D2323
Summary:
There was an old "create_user.php" script but it really was only useful for
creating agents. Provide a more user-friendly script for creating the first
account.
Depends on D278.
Test Plan:
Used 'accountadmin' to create and edit accounts. Read documentation.
Reviewed By: tuomaspelkonen
Reviewers: jungejason, tuomaspelkonen, aran
CC: ccheever, aran, tuomaspelkonen
Differential Revision: 279
