Mediawiki to Confluence migration

2,729 views
2,479 views

Published on

Presentation from Atlassian User Group Hamburg, 6.6.2012.
Topic was migration from Mediawiki and rollout of Confluence in a complex environment with a lot of content.

Published in: Technology
1 Comment
1 Like
Statistics
Notes
No Downloads
Views
Total views
2,729
On SlideShare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
9
Comments
1
Likes
1
Embeds 0
No embeds

No notes for slide

Mediawiki to Confluence migration

  1. 1. Switching to Confluence with 500+ Wiki users Migrating Bigpoint from Mediawiki to Confluence AUGHH user group meeting, 6.6.2012, Nils Hofmeister
  2. 2. Agenda • Before Confluence • The mission • Status quo • Learnings 2
  3. 3. Before Confluence 3
  4. 4. Before Confluence • Time: October 2010 • Bigpoint has >500 employees • There is a bunch of MediaWiki instances (>50) • Some customization 4
  5. 5. Before Confluence 5
  6. 6. Before Confluence We had the wrong tool for the wrong people and it hurt. But barely anybody was aware… Fortunately there were a couple of people interested in replacing our Wiki by Confluence. 6
  7. 7. Before Confluence To justify the costs, we used the following arguments: • Global search • Spaces • Role-based permissions • Connection to Jira • Versioning + concurrency handling • All the plugins • Migration via UWC In late 2010, we got approval. The fight for resources started… 7
  8. 8. The mission 8
  9. 9. The mission Open questions • How to integrate with Bigpoint IT platform? 9
  10. 10. The mission Open questions • How to integrate with Bigpoint IT platform? • • • • Have everything in SVN Wrap Tomcat daemon so it works with monitoring, Ops automation etc Use configuration templates for modified files Setup a staging system 10
  11. 11. The mission Open questions • How to integrate with Bigpoint IT platform? • • • • Have everything in SVN Wrap Tomcat daemon so it works with monitoring, Ops automation etc Use configuration templates for modified files Setup a staging system • Who maintains it? 11
  12. 12. The mission Open questions • How to integrate with Bigpoint IT platform? • • • • Have everything in SVN Wrap Tomcat daemon so it works with monitoring, Ops automation etc Use configuration templates for modified files Setup a staging system • Who maintains it? • My team (Release Engineering) • Right combination of skills and focus, but still… 12
  13. 13. The mission Open questions • How to integrate with Bigpoint IT platform? • • • • Have everything in SVN Wrap Tomcat daemon so it works with monitoring, Ops automation etc Use configuration templates for modified files Setup a staging system • Who maintains it? • My team (Release Engineering) • Right combination of skills and focus, but still… • How exactly will migration happen? 13
  14. 14. The mission Open questions • How to integrate with Bigpoint IT platform? • • • • Have everything in SVN Wrap Tomcat daemon so it works with monitoring, Ops automation etc Use configuration templates for modified files Setup a staging system • Who maintains it? • My team (Release Engineering) • Right combination of skills and focus, but still… • How exactly will migration happen? • • • • First sample spaces as example New “units” go directly to Confluence Migrate Teams step by step using UWC => Soft migration 14
  15. 15. The mission Open questions • How to integrate with Bigpoint IT platform? • • • • Have everything in SVN Wrap Tomcat daemon so it works with monitoring, Ops automation etc Use configuration templates for modified files Setup a staging system • Who maintains it? • My team (Release Engineering) • Right combination of skills and focus, but still… • How exactly will migration happen? • • • • First sample spaces as example New “units” go directly to Confluence Migrate Teams step by step using UWC => Soft migration • What about Kerberos SSO and AD? 15
  16. 16. The mission Kerberos • Not easy to grasp • Hard to deal with when you are not admin • Gave us a lot of trouble in Java context So we used an already existing in-house service: Behold… LoginProxy! 16
  17. 17. The mission 17
  18. 18. The mission Integration • We had a first RC ready in April 2011 • It used LoginProxy for authentication • It used a cronjob + SOAP for AD sync / authorization • We had two blades in place for staging + production: • 2x Quad core, 12 GB RAM, 2x 320 GB HDD, SATA, JBOD • Backup etc via Bigpoint standard mechanisms • Took about 5 man weeks to get everything ready and test it • Central technology teams started using it • Administration was cooperation of Release Engineering + IT Engineering 18
  19. 19. The mission Migration • No interruption of ongoing projects • Long migration timeframe (>6 months) • Lack of acceptance with some users • UWC results very mixed • => More users started noticing Confluence… • Thank god we had a tech writer who could assist with content, support and training 19
  20. 20. The mission Migration • Tracking of wiki migration using Jira • Conversion respecting stakeholder schedules • Mediawikis still exist, but read-only • A lot of training • • • • • • Brown bag meetings Coaching per group Update meetings Confluence space Examples … 20
  21. 21. The mission Result: Success Specs, 06/2012 (14 month later): • 971 users • 152 groups • 152 spaces (without personal) • 19.493 pages created • 34.091 attachments uploaded “You can find our current documentation in Confluence” -Random Bigpoint employee 21
  22. 22. Status quo 22
  23. 23. Status quo • In use worldwide • E.g. Hamburg, Berlin, Malta, San Francisco • Confluence 3.5.13 • Balsamiq • Gliffy • So far 2 custom plugins in development • Custom Jira issue creator • Custom AD synchronizer • Integration with • Jira • Issues macros, shortcut links • Application link • Jenkins • Internal middleware (e.g. mailtool) 23
  24. 24. Status quo 24
  25. 25. Status quo Next big tasks 25
  26. 26. Status quo Next big tasks • Confluence 4 • Delayed to avoid shocking our users with 2 major changes within 1 year • Mixed feelings: markup power users, APIs, coaching,… 26
  27. 27. Status quo Next big tasks • Confluence 4 • Delayed to avoid shocking our users with 2 major changes within 1 year • Mixed feelings: markup power users, APIs, coaching,… • Better Kerberos Integration • Avoid trouble with cached passwords vs. tool integration • Reduces maintenance efforts and reliability 27
  28. 28. Learnings 28
  29. 29. Learnings Acceptance • In general, acceptance was given quickly since • Confluence is fancy • Brings a lot of features • Integrates with Jira nicely 29
  30. 30. Learnings Acceptance • In general, acceptance was given quickly since • Confluence is fancy • Brings a lot of features • Integrates with Jira nicely • Maybe a hard migration would have been easier… • …but we would have had far more haters 30
  31. 31. Learnings Acceptance • In general, acceptance was given quickly since • Confluence is fancy • Brings a lot of features • Integrates with Jira nicely • Maybe a hard migration would have been easier… • …but we would have had far more haters • Remaining haters could be convinced by • Dedicated trainings + support • New features (e.g. heatmap, role-based security,…) • Fast reactions – when we started: immediate changes 31
  32. 32. Learnings Acceptance • In general, acceptance was given quickly since • Confluence is fancy • Brings a lot of features • Integrates with Jira nicely • Maybe a hard migration would have been easier… • …but we would have had far more haters • Remaining haters could be convinced by • Dedicated trainings + support • New features (e.g. heatmap, role-based security,…) • Fast reactions – when we started immediate changes Conclusion: when the field isn’t green, only soft migration works 32
  33. 33. Learnings Costs • When we started about 1,5 persons permanently working on Confluence intro 33
  34. 34. Learnings Costs • When we started about 1,5 persons permanently working on Confluence intro • System integration was much more expensive than expected 34
  35. 35. Learnings Costs • When we started about 1,5 persons permanently working on Confluence intro • System integration was much more expensive than expected • Right now, work on demand • • • • • Bug fixes Plugin development Coaching of new people Changes and extensions Standardization • Basically, 1-2 persons are permanently working on Confluence one way or the other 35
  36. 36. Learnings Costs • When we started about 1,5 persons permanently working on Confluence intro • System integration was much more expensive than expected • Right now, work on demand • • • • • Bug fixes Plugin development Coaching of new people Changes and extensions Standardization • Basically, 1-2 persons are permanently working on Confluence one way or the other Conclusion: 2 fulltime persons needed for a Confluence of our size and usage scenario: a DevOps guy and a workflow person 36
  37. 37. Learnings Enterprisy requirements • Authentication and authorization requires customization 37
  38. 38. Learnings Enterprisy requirements • Authentication and authorization requires customization • Certain IT requirements hard to address • Replication • Failover • Automated deployment 38
  39. 39. Learnings Enterprisy requirements • Authentication and authorization requires customization • Certain IT requirements hard to address • Replication • Failover • Automated deployment • Some features are not yet convenient enough • • • • • Bulk attachment upload Easy update of attachments (e.g. excel files) Default groups for new users Notification email templates … 39
  40. 40. Learnings Enterprisy requirements • Authentication and authorization requires customization • Certain IT requirements hard to address • Replication • Failover • Automated deployment • Some features are not yet convenient enough • • • • • Bulk attachment upload Easy update of attachments (e.g. excel files) Default groups for new users Notification email templates … Conclusion: If you want to customize Confluence significantly, you will need admin and Java dev skills. 40
  41. 41. Summary 41
  42. 42. Summary •The good • Soft migration via UWC worked for us • Users were happy quickly • The possibilities are awesome •The bad • The frontend is fancy, maintenance can be weird •The ugly • It costs quite some manpower for serious operation • It needs continuous effort for acceptance • You need skilled, hard to find people for this 42
  43. 43. Summary If you want to operate a serious Confluence instance, you need manpower. But you get the best possible documentation system I know. 43
  44. 44. Contact us Bigpoint GmbH Nils Hofmeister Lead Integration Architect Drehbahn 47-48 20354 Hamburg Germany Tel +49 40.88 14 13 - 0 Fax +49 40.88 14 13 - 11 nhofmeister@bigpoint.net www.bigpoint.net Bigpoint Inc. Bigpoint Distribuição de Entretenimento Online Ltda. 500 Howard Street Suite 300 San Francisco, CA 94105 Av. Brig. Faria Lima 3729 cj. 528 04538-905 São Paulo Brazil Bigpoint GmbH Bigpoint International Services Limited Alexanderstraße 5 10178 Berlin Germany 1 Villa Zimmermann Ta’Xbiex Terrace XBX 1035 Ta’Xbiex Malta Find us on 44
  45. 45. Bigpoint GmbH First name, last name Title Drehbahn 47-48 20354 Hamburg Germany Tel +49 40.88 14 13 - 0 Fax +49 40.88 14 13 - 11 info@bigpoint.net www.bigpoint.net Find us on 45

×