Home / Projects / Orchestrata Ultimate / Decisions / ADR index / ADR-0119

ADR-0119: dr-hannah connects only to the prod orchgraph memgraph cluster; data stranded in another memgraph moves to prod

field	value
Status	Accepted — Accepted (ruling, Mohamed, 2026-10-20)
Date	2026-10-20
Register	factory (`docs/adr`, 4-digit ids)
Source	`docs/adr/0119-dr-hannah-prod-memgraph-cluster-only.md` — permalink at `37e1e95ba` (unchanged on master since that commit)
Wiki register entry	ADR-0119 (wiki status: Accepted)

← ADR-0118 · ADR-0120 →

The record

Status: Accepted (ruling, Mohamed, 2026-10-20) · Date: 2026-10-20

Context

dr-hannah is the code-assistant API (uvicorn dr_hannah.api:app, host port 9100). Verified live (ss, PID 1381) and from config (/opt/dr-hannah/env/prod.env):

Its data store is the prod orchgraph HA cluster (default incus project): DR_HANNAH_MEMGRAPH_HOSTS=10.53.1.128,10.53.1.123 (data nodes, port 7687, database memgraph), with live sockets on orchgraph-data-1:7687 and orchgraph-coord-3:7691 (routing).
Its data already lives on the prod cluster's memgraph database (19,887 nodes): Entity 12,150, CodestralChunk 6,859, Hannah 753, Module 116, ChatMessage 6, ChatSession 3.
The "other memgraph" — codetools-memgraph (orchestrata incus project, 10.222.9.240:7687, standalone) — was verified to hold 0 nodes (one on-disk database, empty). Nothing of dr-hannah's is stranded there.
A stray ambient MEMGRAPH_URI=bolt://10.222.9.240:7687 sits in dr-hannah's process environment, but dr-hannah's code does not reference MEMGRAPH_URI (grep across src/) — it is inert.
Open item: codetools currently target codetools-memgraph via CODETOOLS_MEMGRAPH_URI (/opt/codetools/current/env/test.env), so "the same memgraph as code tools" is not yet true of code tools themselves; that repoint is a separate, tracked follow-up.

Decision (Mohamed, 2026-10-20, verbatim)

"dr-hannah should connect to the ssme memgraph as prodcustion orchestrata and code tools, if it has data in the other memgraph it must be moved to the prof memgraph cluster"

Implemented as: dr-hannah connects only to the prod orchgraph memgraph cluster (the cluster production orchestrata uses). If dr-hannah data is found in any other memgraph, that data is migrated to the prod cluster. The prod cluster is canonical for dr-hannah.

Alternatives rejected

Point dr-hannah at codetools-memgraph — rejected: the ruling names the prod memgraph cluster as the destination of record.
Keep dual endpoints — rejected: two stores for one dataset re-creates the split-brain the ruling eliminates.

Consequences

Verified at ruling time: nothing to move — the other memgraph holds 0 nodes, and dr-hannah's data is already on the prod cluster's memgraph database. dr-hannah's prod.env already complies; the stray ambient MEMGRAPH_URI is inert and documented as such.
Codetools' CODETOOLS_MEMGRAPH_URI still points at codetools-memgraph; repointing codetools to the prod cluster is the tracked follow-up that completes the ruling's "same memgraph as code tools" clause, and is the precondition for retiring codetools-memgraph.

Provenance

Mohamed, in-session, 2026-10-20 (quote above, verbatim). Evidence: live ss of PID 1381; Bolt queries against 10.53.1.128:7687 (db memgraph) and 10.222.9.240:7687; /opt/dr-hannah/env/prod.env.

No other ADR is named in this record.

Rulings cited

None.

Specs cited

None (or no spec page exists for the ones named).