444c4d5999
* LDAP: Divide the requests Active Directory does indeed have a limitation with 1000 results per search (default of course). However, that limitation can be workaround with the pagination search feature, meaning `pagination` number is how many times LDAP compatible server will be requested by the client with specified amount of users (like 1000). That feature already embeded with LDAP compatible client (including our `go-ldap`). But slapd server has by default stricter settings. First, limitation is not 1000 but 500, second, pagination workaround presumably (information about it a bit scarce and I still not sure on some of the details from my own testing) cannot be workaround with pagination feature. See https://www.openldap.org/doc/admin24/limits.html https://serverfault.com/questions/328671/paging-using-ldapsearch hashicorp/vault#4162 - not sure why they were hitting the limit in the first place, since `go-ldap` doesn't have one by default. But, given all that, for me `ldapsearch` command with same request as with `go-ldap` still returns more then 500 results, it can even return as much as 10500 items (probably more). So either there is some differences with implementation of the LDAP search between `go-ldap` module and `ldapsearch` or I am missing a step :/. In the wild (see serverfault link), apparently, people still hitting that limitation even with `ldapsearch`, so it still seems to be an issue. But, nevertheless, I'm still confused by this incoherence. To workaround it, I divide the request by no more then 500 items per search
Grafana backend codebase
The code styleguide and brief description of the architecture
On going refactorings.
These issues are not something we want to address all at once but something we will improve over time. Since Grafana is released at a regular schedule the prefer approuch is to do this in batches. Not only is it easier to review, it also reduces the risk of conflicts when cherry-picking fixes from master to release branches. Changes that spawn multiple locations are therefore prefered in the end of the release cycle since we make fewer patch releases in the end of the cycle.
Global state
Global state makes testing and debugging software harder and its something we want to avoid when possible.
Unfortunately, there is quite a lot of global state in Grafana. The way we want to migrate away from this
is to use the inject package to wire up all dependencies either in pkg/cmd/grafana-server/main.go or
self registering using registry.RegisterService ex https://github.com/grafana/grafana/blob/master/pkg/services/cleanup/cleanup.go#L25
Reduce the use of the init() function
Should only be used to register services/implementations.
Settings refactoring
The plan is to move all settings to from package level vars in settings package to the setting.Cfg struct. To access the settings services/components can inject this setting.Cfg struct.
Reduce the use of Goconvey
We want to migrated away from using Goconvey and use stdlib testing as its the most common approuch in the GO community and we think it will make it easier for new contributors. Read more about how we want to write tests in the ARCHITECTURE.MD docs.
Sqlstore refactoring
The sqlstore handlers all use a global xorm engine variable. This should be refactored to use the Sqlstore instance.
Avoid global HTTP Handler functions
HTTP handlers should be refactored to so the handler methods are on the HttpServer instance or a more detailed handler struct. E.g (AuthHandler). This way they get access to HttpServer service dependencies (& Cfg object) and can avoid global state
Date comparison
Newly introduced date columns in the database should be stored as epochs if date comparison is required. This permits to have a unifed approach for comparing dates against all the supported databases instead of handling seperately each one of them. In addition to this, by comparing epochs error pruning transformations from/to other time zones are no more needed.
Dependency management
Documented in UPDRAGING_DEPENDENCIES.md.