Come for the products,
stay for the community

The Atlassian Community can help you and your team get more value out of Atlassian products and practices.

Atlassian Community about banner
4,367,525
Community Members
 
Community Events
168
Community Groups

How large should an external SQL Server database be for Crucible?

Edited

This question was asked and answered for Stash but I don't see one for Crucible.

2 answers

1 accepted

5 votes
Answer accepted
Conor Atlassian Team Mar 17, 2013

Sergey is correct. In FishEye/Crucible, the SQL database is primarily used for Crucible data. The FishEye data is stored locally in files using an embedded database and also in Lucene indexes.

Given that, the sizing of the SQL database will be driven mostly by the number of reviews, the number of file revisions in each review and the number of comments made on each review.

A ballpark figure might be to allow:

  • 1k per file revision
  • 512 byes per review comment
  • 2k per review

Given a review average of 30 file revisions and 20 comments, you'd arrive at 42k per review. So for 20k reviews, you'd be looking at 800 Megabytes.

Alternatively, if you were to create 10 reviews per week, your Crucible data usage will be around 22 MBytes per year.

This is, of course, a Fermi calculation of sorts and your actual usage is going to depend a lot on the number of reviews you conduct, the complexity of the reviews and how much discussion there is within the reviews.

Here's some info from a live instance (numbers are average row size in bytes). Largest table by row count is cru_fr_detail, btw (10x larger than next largest, cru_revision).

| cru_base_star_model           |   309 |
| cru_changeset_comment         | 16384 |
| cru_comment                   |   251 |
| cru_comment_field             |   116 |
| cru_comment_read_status       |    50 |
| cru_committer_user_mapping    |   122 |
| cru_content_root              |   204 |
| cru_field                     |    78 |
| cru_file_read_status          |    47 |
| cru_fr_detail                 |    71 |
| cru_frx                       |    62 |
| cru_frx_comment               |    42 |
| cru_frx_revision              |    49 |
| cru_inline_comment            |    31 |
| cru_inline_comment_to_frx_rev |    39 |
| cru_logitem                   |   164 |
| cru_metric_definition         | 16384 |
| cru_notification              |    56 |
| cru_patch                     |   348 |
| cru_patch_revision            |    50 |
| cru_perm_scheme               |   218 |
| cru_proj_allowed_grp          |   212 |
| cru_proj_default_grp          |   655 |
| cru_proj_default_reviewer     | 16384 |
| cru_project                   |   218 |
| cru_ps_all_user               |    72 |
| cru_ps_group                  |    75 |
| cru_ps_review_role            |    76 |
| cru_recently_visited          |   148 |
| cru_recipient                 |    64 |
| cru_review                    |   663 |
| cru_review_comment            |    37 |
| cru_review_participant        |    51 |
| cru_revision                  |   120 |
| cru_revpermaid                |    60 |
| cru_state_change              |    52 |
| cru_stored_path               |   151 |
| cru_upload_item               |   106 |
| cru_user                      |   137 |
| cru_user_profile              |   142 |
| cru_version                   |  8192 |

0 votes

Hi there,

I found this question about database size for Stash, is this the one you are refering to? In that question Michael's calculation is very smart, but keep in mind that it should be considered a guideline and not a rule.

I believe we could use the same logic for Crucible. However, since Crucible and Fisheye use the same database, it should be included in the calculation formula.

So, I'd say 100MB + ((total number of commits across repos + total number of reviews per repo) / 2500) MB

Using the same calculation basis, if you have 20 repositories in Fisheye with an average of 25,000 commits and 20,000 reviews you'd need 100Mb + (20 * (25,000 + 20,000) / 2500) MB = 460MB.

I hope this helps.

Cheers

This doesn't look right -- primarily because SQL DB is not used to store commit metadata. But since schema is undocumented, I cannot offer any better formula, perhaps someone from Crucible team can.

Suggest an answer

Log in or Sign up to answer
TAGS

Atlassian Community Events