Professional Services EM Scoping - Migrations

Describes the processes for scoping migrations.

Info for Sales - positioning migrations to SaaS

Migration Scoping Details

This page describes scoping migrations from GitLab, Bitbucket Server or GitHub (Enterprise or .com) to a destination GitLab instance (Self-Managed or SaaS). The migrations typically utilize our Congregate automation tool. Migrations from other SCM systems and non-GitLab CI/CD migrations are out of scope for this migration tooling and must be scoped separately.

Using the services calculator, an SA or CSM/CSM can create scoping issue, and work with an Engagement Manager to iterate and refine the services estimate for a customer. In this issue, we have included additional context to the SCM Migration Scoping Questions, which can be previewed below

SCM Migration Scoping Question Customer Response Example response Rationale for asking
Source SCM System(s) to-do Bitbucket Server, GitLab Self-Managed GitLab PS has automation to facilitate migration for the most popular source systems. Need to know what systems the data is coming from to accurately scope the time to migrate.
Destination GitLab deployment (SaaS, Self-Managed(HA), Self-Managed(single node)) to-do SaaS We ask this to ensure the destination system deployment is strong enough to handle the throughput of data that will be required during migration.
Source GitLab version (must be 2 behind latest) to-do 13.4 Migration services leverage many GitLab APIs including our project import/export API. There are specific compatability guidelines documented here
Destination GitLab version (must be within 2 minor versions of source) to-do 14.1 Migration services leverage many GitLab APIs including our project import/export API. There are specific compatability guidelines documented here
Total number of Users** to-do BB = 225, GLSM = 775 Migrating users is a pre-requisite step to migrating the data to ensure data elements are associated properly. This is a discrete task in migration engagements and must be scoped with number of users as an input.
Total portfolios (w/ stakeholder rep) to-do 6 We use total portfolios as a proxy metric to identify how much coordination will be required during migration. Each portfolio leader needs to understand and buy into the migration process for things to go smoothly. This coordination time is built into the migration engagement.
Total number of git repositories to-do BB = 1,234; GL = 4321 We need to know the total number of repositories to be migrated in order to estimate the number of days of effort. Note: A bitbucket project can contain multiple git repositories. We need the total number of repos.
Total >5GB GitLab repositories to-do BB N/A, GL = 3 Only applicable for GitLab self-managed migrations to GitLab.com; repos >5GB must be manually migrated due to a 5 GB Cloudflare limitation. Note that we only care about repo size, not the total project size.
CI/CD System(s) to-do Jenkins We need to know what CI/CD systems are being used to estimate how much it will take to repoint those pipeline jobs to the new SCM system to help the customer resume IT operations.
Total ci/cd jobs? (CI/CD jobs will need cut-over even if not migrated) to-do 4567 The engagement could include repointing CI/CD jobs back to a source repository. If this is the case, we will need to know how many jobs need to be reconfigured.
Typical registry size to-do 159MB If registry sizes are unusually big, it could affect the speed of migration. Note: this question only applies to migrations where GitLab is a source system.
SSO Identity Provider to-do Auth0 We want to make sure this is already in place prior to migration as it is a foundational to the success of a migration engagement. See here for a full list of supported Identity Providers
Source system OS and Version to-do Ubuntu v21.10 If an upgrade of the source system is needed/included prior to the migration, we want to be sure the OS does not need to be upgraded by the customer to support the new version of GitLab. See installation requirements for more details.
Source system DB version to-do PostgreSQL 13.0 If an upgrade of the source system is needed/included prior to the migration, we want to be sure the DB does not need to be upgraded by the customer to support the new version of GitLab. See installation requirements for more details.

Notes:

  1. Source Code Management (SCM) Migrations do not include secrets mediation.
  2. SCM migrations do not include transitioning CI/CD functionality to use GitLab CI.

GitLab Self-Managed to GitLab.com Migration Notes

  • There are specific cases where migrating to SaaS might not be the best recommendation. Here are some considerations when working with the customer to decide on whether or not SaaS is the best solution:
    1. Is the customer using something other than GitLab CI? If so, are they able to establish connectivity between their CI tool and GitLab.com? There are occassionally issues that prevent them from using GitLab and their CI/CD tools, i.e. the customer is not able to convince their network security team to adjust firewall configurations to be able to use their CI/CD tool behind their firewall with GitLab SaaS.
    2. If the customer has a very large number of projects, this could become a change management challenge as well as be cost prohibitive to do the migration. e.g. for one customer recently, who has been having performance issues in their self-managed instance, it was going to take $400k and 150+ days to migrate them to SaaS (1600 users and 20k+ projects). GitLab Dedicated might be a better option for some customers who are using Ultimate.
    3. The customer may be using some self-managed features that they are insistent on continuing to use that aren’t available on SaaS (e.g. server hooks). See Differences Between Self-Managed and SaaS for more information.
    4. The customer may have very large GitLab projects that would be difficult to migrate to SaaS, and they would need to pay for additional storage consumption in order to use SaaS; plus there are known git and CI performance issues when customers have very large monorepos.
  • Migrating from Self-Managed to SaaS has additional information, rough order magnitude estimates, and common customer questions
  • see Congregate Features Matrix to see a comparison of GitLab ui migration vs Congregate migration and gain a better understanding of which features get migrated and which don’t.
  • Migrations from one GitLab instance to another using Congregate require that the group and project structure of the source system is retained during migration.
  • scoping uses our standard Automated Migration PS Engagement Estimate Template. For very small migrations with less than 30 repositories, use the Manual Migration Estimate Template to compare the cost of a manual migration to the cost of automated migration
  • SaaS Discovery, SSO configuration and Security configuration activities are typically added to migrations to SaaS. See Services Calculator activities for more details:
  • see TEMPLATE Professional Services Presentation for a summary of our migration approach and which features get migrated or not. This also includes a description of our SaaS Discovery, SSO configuration and Security configuration activities, which we often add for new SaaS customers.

GitLab Self-Managed to GitLab Self-Managed

  • You can use Congregate but you also have other options at your disposal:
    • Backup and restore using rake-tasks. This enables you to restore all of the data from one instance of GitLab to another. This can be used to shorten the migration period, but will cause complete downtime as the source instance will need to be locked during the time it takes to backup. Also, this could take many hours to days depedening on the size of the data.
    • Geo Replication. This method allows you to setup the new GitLab instance as a Geo secondary. The data will sync over the course of time. Then you can perform a failover to make the new instance primary. This tends to be more complex and requires a more specialized skillset than the backup/restore approach.

GitLab Self-Managed to GitLab Dedicated Instance

  • As with self-managed to self-managed migrations, self-managed to dedicated instance migrations can be performed using Congregate, a backup and restore approach, or Geo replication. If the Congregate option is used, projects can be migrated in waves of up to 500.
  • Migrating groups with projects is a Beta feature could also be used to migrate to dedicated instance. Dedicated instance needs to be on GitLab version 15.10 or later. Please review linked documentation to understand limits and requirements of this method.

GitHub Source

GitHub Enterprise to GitLab Self-Managed

  • This is one of our strongest combinations of source and destination systems. We have migrated 7,000+ projects per wave in our best case.
  • If minimizing the transition period is a concern, we will want to maximize the number of projects per wave.
    • To do this, we need to have full control over the GitHub API Rate limit setting.
    • We also want to ensure there is a strong enough network connection between the GHE and GitLab SM instance to support moving the data. We don’t have a hard number, but ask them about moving hundreds of GB over a few hours across lots of connections. Their network team should be able to answer.
  • If the customer wants to reorganize their groups/project structure, we can support this (see below common customer requests).
  • Make sure the customer acknowledges that they need to have email addresses public, not private for any migration to work properly.
  • GitLab self managed should be on 13.7+ or later to take advantage of an API change that allows us to specify a specific github hostname in the import call (rather than defaulting to github.com).

GitHub Enterprise to GitLab SaaS

  • The destination system rate limit will be the limiting factor in these migrations. We have no control over gitlab.com rate limit so these will need to be scoped at 200-300 per wave.
  • GitLab self managed should be on 13.7+ or later to take advantage of an API change that allows us to specify a specific github hostname in the import call (rather than defaulting to github.com).
  • Make sure the customer acknowledges that they need to have email addresses public, not private for any migration to work properly.

Github.com to GitLab Self-Managed

  • The Github.com API rate limit is very low per user so we workaround by using multiple userIDs provided by the customer.
  • The total projects migrated per day should not exceed 200 until we deliver this to a customer at scale. We’ve done it at small scale in terms of number of projects. But some of those projects were gigantic (70K+ Pull Requests).
  • Make sure the customer acknowledges that they need to have email addresses public, not private for any migration to work properly.

Github.com to GitLab SaaS

  • We are limited by both the source and destination system rate limits.
  • Make sure the customer acknowledges that they need to have email addresses public, not private for any migration to work properly.

Bitbucket Source

_Note: A project on bitbucket is equivalent to a GitLab group. A Repository on Bitbucket is equivalent to a GitLab project.*

Bitbucket Server to GitLab Self-Managed

  • Theoretically, migrations with this pair of source/destination should be able to be scaled as high as GHE to GL Self-Managed. Its safe to increase the projects per wave to 1,000.
  • If minimizing the transition period is a concern, we will want to maximize the number of projects per wave.
    • To do this, we need to have full control over the BitBucket API Rate limit setting.
    • We also want to ensure there is a strong enough network connection between the BitBucket Server and GitLab SM instance to support moving the data. We don’t have a hard number, but ask them about moving hundreds of GB over a few hours across lots of connections. Their network team should be able to answer.
  • If the customer wants to reorganize their groups/project structure, we can support this (see below common customer requests).

Bitbucket Cloud to GitLab Self-Managed

  • GitLab does not have an API to initiate an import from bitbucket cloud currently. Automated migrations are not possible.
  • Can position a teach a customer to fish advisory approach that uses the BB cloud import UI to help with importing.

Bitbucket Server to GitLab SaaS

  • We will be limited by the rate limit on GitLab SaaS, which is why our scoping is low (e.g. 200 projects per wave)

Bitbucket Cloud to GitLab SaaS

  • GitLab does not have an API to initiate an import from bitbucket cloud currently. Automated migrations are not possible.
  • Can position a teach a customer to fish advisory approach that uses the BB cloud import UI to help with importing.

Team Foundation Server (TFS)/Azure DevOps (ADO) to GitLab

Azure DevOps (formerly named Team Foundation Server) contains more than a source code repository, so additional questions need to be asked while scoping out an ADO migration. Five common components in ADO are the following:

  • Azure Boards: Set of Agile tools to plan, structure and manage work through a backlog. Equivalent to GitLab Issues or JIRA tickets.
  • Azure Repos: Version control system. Supports Git or Team Foundation Version Control (TFVC).
  • Azure Pipelines: Build and release services to support continuous integration and delivery of an application. Equivalent to GitLab CI/CD.
  • Azure Test Plans: Several tools to test an application, including manual/exploratory testing and continuous testing.
  • Azure Artifacts: Service to store and share packages such as Maven, npm, NuGet, generic and others from public and private sources. Equivalent to GitLab Package Registry.
Question Answer Sample Answer Rationale for asking
General
Is it SaaS or self-hosted? Azure DevOps (SaaS) If ADO Server (self-hosted), then we need to know software version and other details (i.e. SQL version etc), this will help to pick the right migration approach.
How many organizations do you have that need to be migrated? 1 Answer to this question will help to understand the current state, amount of required effort and contribute to future architecture of GitLab.
How many projects per organization? 500 Answer to this questions will help to estimate amount of time and approach for migration.
What is the ratio project/repo? 1:1 Based on answer to this question additional advisory services to refactor/restructure might be required
Azure Boards
Are you using workitems in ADO? Yes If workitems need to be retained, then additional migration and/or advisory activities need to be added to the engagement. Note: there are features in ADO workitems (e.g. custom fields and workflows) that are not supported by GitLab issues. During scoping, make sure the customer is aware of these differences.
What template do you use? Scrum, Agile, SCCM To understand the current state and ensure GitLab’s Issues support all the capabilities.
What customizations do you use in your boards? Extra swimming lanes, custom fields etc Based on this answer we may require additional advisory services to discuss what is possible / not possible and propose alternatives.
Azure Repos
Are you using Git or TFVC for your SCM? Git This will influence how we interact with the ADO server and determine if a conversion to Git is necessary
How many code repositories? 500 This will affect the decision on general approaches listed above
Are you using branches in your code repositories? Yes If the answer is no for some of the repositories, the customer will have to choose between converting specific folders in the repository to branches to retain history or accept a flat file migration of the repository with no history
Do you need to retain history in your code repositories? Yes If the answer is no and the customer is using ADO, then we do not need to convert the repos to Git.
Azure Pipelines
Are you using ADO to build your software? Yes This will add CI/CD consulting/transformation activities to the engagement if the answer is yes
How many builds/build templates are used per code repository? 1 This is a gauge of complexity. Sometimes a code repository can contain several different build definitions
Are multiple solution (.sln) or project (.csproj) files building the same packages? Yes If yes, this can require advisory services to refactor to those solution/project files to work with GitLab pipelines
How many build servers do you use? 1 We usually convert a build server to something more ephemeral so the more build servers in use, the more development is required to transition them to something more ephemeral
Are there any specific flags used in your build process? If so, what? Yes This shows the customer is measuring this data to be used during the transition process.
Do you use classic releases? No If the answer is yes then additional effort might be required to educate on YAML (pipeline as code) concept
Do you use UI build definitions? No If yes then this will add CI/CD consulting/transformation activities
Do you use YAML for all your definitions (build / release)? Yes If “Yes” is the answer to this question then SOW might include additional effort and advisory services (i.e. CI/CD consulting/transformation activities)
Do you use extensions from the marketplace that are part of your build/test/release process? Provide the list of extensions (if the answer is “Yes”) Yes. ARM-ttk extension for linting templates. If yes then we need to ensure that we have equivalent in GitLab for this (or propose alternative solution)
What type of agents do you use? Self-Hosted, Microsoft-hosted (i.e. windows-latest, ubuntu-latest etc) Based on this answers we will make a decision what runner architecture to propose and make sure GitLab supports it and how much effort it requires to achieve similar state
Do they have special requirements (i.e. network connectivity, required components for build/test/release process)? Yes, need Visual Studio 2017 capabilities (i.e. legacy software) Based on the answer we will make a decision if custom runner images are needed (i.e. to install capabilities)
Are there any plans in the future to use/add anything special in the CI/CD? GPU optimized agent Based on the answer we will make a decision on GitLab Runners to avoid potential limits in the future
Azure Test Plans
Do you use test plans and do you need similar functionality in GitLab? Yes Based on the answer we need to propose equivalent feature in GitLab
Azure Artifacts
Do you use artifact feeds? Yes To understand the current state
What type of feeds? Maven, npm, NuGet To understand the current state and make ensure GitLab has the same capabilities (see GitLab Package Registry)
Do you use build/pipeline artifacts? Yes
What is the retention policy of the packages and do you need to migrate them? 365 days, no migration needed To estimate additional effort for migration or/and advisory service to retain existing ADO feed refactoring application to use GitLab Package Registry
Integrations
Are there any external tools/applications tied to your ADO server? Yes. We use an in-house tool that pulls from TFS daily for gathering metrics If the answer is yes, additional activities will need to be added to the SOW to accommodate transitioning those tools to pull from Git instead.

Other git based SCMs

  • We can support these customers by using the “bare git” method of migration. This is done through the Import repo by URL UI or command line using git push -u.
  • The customer should provide a list of git urls to iterate over to support the migration.
  • Data elements outside of the git envelope (e.g. pull request comments, user membership, etc.) will not be migrated. Only git data elements (e.g. branches, commits, files, tags, etc.) will be migrated.

Non-git based SCMs

  • For sources like CSV, ClearCase, SVN, TFVC we can support non automated migration but there is risk that should be factored into these engagements as we dont have deep experience with these source systems. Consider working with a channel partner to find expertise in these source systems.
  • Ensure the customer has researched and thought about how their projects will be organized on the destination GitLab system. Often times during the initial transformation from non-git SCM to git scm, there will be splitting of a single trunk of source code to multiple git repositories.

Common customer requests

1. Can we reorganize our projects and groups during migration?

  • In congregate, we can do this for source systems of Github Enterprise and Bitbucket. The customer must provide a .csv file with source project url and destination parent path to let us know where the projects will land on the destination system. Also the customer or we must create the group structure prior to a migration with this reorganization involved. There should be additional time added for new group hierarcy creation to the scope.
  • We strongly recommend against this for GitLab as the source and destination systems. This seems counterintuitive, but because GitLab has lots of other data outside of the repository structure that sits in the project/group structure that needs to be maintained, the reorganization step should not happen during the migration. If a customer MUST do this, recommend doing it as a pre or post migration step using the UI or API to coordinate the local moves of projects and groups.
    • This can be positioned as additional days of migration advisory time. We need to determine what this is a function of to have objective scoping.

2. We’re on 13.5 (or some old version), can we still migrate?

  • The more of a difference between the source and destination software versions, the higher likelihood there is for data integrity to be lost during migration. The team who maintains the import functionality documented that the importers work when source and destination are no more than 2 minor releases apart from each other. For this reason we include upgrade services to help customers get their source instance to be matching or no less than 2 minor versions from the destination software version. Only in extreme circumstances will we consider exceptions to this guidance.

Other Resources

  • See the publicly available Migration Toolkit to help understand the steps taken during delivery of migration services.
  • The Customer folder of the migration template project contains useful documentation to help a customer prepare for their migration.
  • The Migrating from Self-Managed to SaaS handbook page has details specific to migrating to gitlab.com.

Migrating from Self-Managed GitLab to GitLab.com
Things to know when helping customers migrate to GitLab.com.
Last modified July 10, 2024: Fix broken links and spelling (680a0bc8)