Closed andrew-jameson closed 4 weeks ago
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 90.65%. Comparing base (
25b762b
) to head (5b32245
). Report is 2 commits behind head on develop.
@andrew-jameson @lfrohlich @ttran-hub -- below is my feedback related to this ADR, most of these are geared toward trying to anticipate questions from our product and security teams:
diagram-related
- recommend adding the ATO boundary to make it more clear that this solution will be hosted outside of it
Andrew to denote this in diagram
- where are the other tools (PLG and associated) in this diagram?
under clamav in diagram
- Does the triangle represent load balancers?
triangle represents the proxies
cost-related
- Can we lay out the monthly costs associated with hosting Sentry in Cloud.gov compared to using Sentry's SaaS offering?
$26/mo for Sentry SaaS vs $130/mo * 8GB-10GB if we hosted Sentry in our cloud.gov environment
reference: https://sentry.io/pricing/
- Do we have a preliminary sense of what might increase these monthly costs over time (on the Sentry SaaS side)? What additional expenses (e.g., scaling costs) might be incurred as part of this solution?
could go up to $80/mo (Sentry pricing Business tier) if we need custom dashboards. it's too early to decide this.
security-related
- Does self-hosting mean that the other tools (PLG stack and associated) will be hosted in Cloud.gov? If so, what are the associated costs, and how do these compare to other hosting options? If not, where will this be hosted, and do we have visibility into its security compliance standards (e.g., encryption, access controls)?
PLG to be hosted in cloud.gov requires about 10GB (*$130/mo), but not sure just yet.
- What security and compliance standards does the Sentry SaaS adhere to (e.g., SOC 2, FedRAMP, HIPAA)?
SOC 2 (reference: https://sentry.io/security/#third-party-audit)
- Is the data stored in Sentry’s SaaS encrypted in transit and at rest? If so, what encryption standards are used? https://sentry.io/security/#data-security-and-privacy
- Will any PII or other sensitive data be stored outside of our ATO boundary? If so, please describe how it will be protected.
access control-related
- Can you describe the types of information to be exchanged between our system’s boundary and Sentry’s SaaS, including what logs, metrics, or other data will be sent or accessed? And what info will not be exchanged?
- How will sys admins access this information, and will it require accessing data outside of our system’s boundary diagram?
see above
- If information flows back into our boundary diagram, how are we ensuring that this process doesn’t create a new vector for potential attacks or data leaks?
mostly outbound communication (not sure about inbound yet - TBD spike)
- What access controls and monitoring mechanisms will be in-place for these external systems to detect and prevent unauthorized access?
must go through nginx to reach our backend. if sentry compromised... incident response (https://sentry.io/security/#corporate-security) we may need our own IR plan.
@lfrohlich TDP IPT (external) is next Wednesday -- dev plans to present this. cc: @andrew-jameson @vlasse86 @ttran-hub
The additional info above is helpful but @andrew-jameson could you summarize what the expected cost would be, or the range of costs with all components considered?
@lfrohlich On low end, total would be $806/mo.
Per @elipe17, the range for PLG is approx 6-10GB of RAM total to cover all 3 spaces; above quote is for 6GB. Sentry would be $26/mo flat fee.
We need to actually calculate memory requirements for PLG. My assumption is Loki: 2-4GB, Prometheus: 2-4GB (calculator), Grafana: 1GB, 3 PG exporters: 3 24MB, 6 Backend Promtails: 6 64MB.
The additional info above is helpful but @andrew-jameson could you summarize what the expected cost would be, or the range of costs with all components considered?
@lfrohlich On low end, total would be $806/mo.
Per @elipe17, the range for PLG is approx 6-10GB of RAM total to cover all 3 spaces; above quote is for 6GB. Sentry would be $26/mo flat fee.
We need to actually calculate memory requirements for PLG. My assumption is Loki: 2-4GB, Prometheus: 2-4GB (calculator), Grafana: 1GB, 3 PG exporters: 3 24MB, 6 Backend Promtails: 6 64MB.
Please see my updated memory requirements here.
Please see my updated memory requirements here.
Thank you @elipe17! New low-end cost would be $416/mo @lfrohlich
Please see my updated memory requirements here.
Thank you @elipe17! New low-end cost would be $416/mo @lfrohlich
low end: 3GB x $130 (RAM cost per GB)+ $26 (Sentry) = $416/mo high end (doubling GB estimate for budgeting purposes): 6GB x $130 + $26 = $806/mo
@lfrohlich
Summary of Changes
Provide a brief summary of changes Pull request closes #3199
Please also see #3206