A user clicks a button. The request stalls. Ten seconds of silence. They click the fallback button. Now two jobs are running against a single intention. You end up with duplicate side effects, double charges, and a data mess that eats your afternoon.
This is not a frontend bug. A disabled button or a debounce timer in React will not save you. The first request was already in flight. The network simply swallowed the response. If your backend treats every incoming request as a brand new instruction, retries become liabilities. You need to fix this in your API design and your database schema.
The solution starts with a simple structural split.
Split Jobs from Attempts
Think of a job as the durable record of what the user wants. It captures the owner, the parameters, the target provider, and the exact intent. An attempt is a specific try at fulfilling that intent.
Picture a print shop. You hand over a file and they give you ticket #45. That ticket is the job. The shop tries the inkjet printer. It jams. That is attempt one. They move the file to the laser printer. That is attempt two. Throughout the process, ticket #45 never changes. If the shop issued a new ticket for every printer they tried, you would pay three times and receive three unwanted copies.
Your database should mirror this. One table holds jobs. Another table holds attempts. The job row stays constant while the attempts accumulate beneath it.
This separation gives you control. It also gives you a place to attach an idempotency key that survives network blips.
Require an Idempotency Key on Every Job
Every POST request that creates a job must carry a unique idempotency key. This key belongs to the user, not the session. Combine the owner ID and the key, then enforce a unique database constraint across those two columns.
Why a database constraint? Because checking for existence in application code before inserting is a race condition waiting to happen. Two identical requests can slip through the same microsecond gap. Let the database be the enforcer. If a user sends the same owner ID and key twice, the second request catches the unique violation and you return the existing job. Both requests get the same job ID. No duplicate work starts.
Be strict about scope. If someone reuses the key but changes the input payload, return a conflict. The idempotency key must bind to an exact intention, not just the user. Same key with different input means the client is confused, and your system should reject it rather than guess.
Protect State Transitions
An attempt is a state transition, not a new job. Your API must refuse to spawn a fresh attempt if a previous attempt is still hanging in a starting or unknown state.
Timeouts are the reason. When a provider request times out, the client sees failure, but the server-side process might still be alive. The GPU cluster could still be churning on your inference request. The container might still be writing to blob storage. If you mark the timed-out attempt as failed and immediately fire a second attempt, you are gambling with duplicate side effects.
Treat a timeout as an unknown state, not a failed one. Block new attempts until the earlier one reaches a terminal state or is explicitly cancelled by an out-of-band process. This pause is uncomfortable. It forces the user to wait. It also prevents the chaos of two workers mutating the same downstream resources.
Resolve Races with Compare-and-Swap
The hardest problems show up when multiple attempts finish. Maybe your system fired attempt one against the primary provider. After ten seconds of silence, it fired attempt two against the fallback. Now both attempts are done. You cannot let both write their results to the same job row.
Use compare-and-swap logic. Add a version number to the job row. When an attempt finishes, it runs an update with conditions:
- The current version must match what the attempt read at the start.
- No other attempt must have already claimed the result slot.
- If both pass, write the result and increment the version.
In SQL terms, that looks like an update statement with a WHERE id = $1 AND version = $2 AND completed_by IS NULL. If the update returns zero rows, another attempt already won. The late arrival must be ignored. Drop its result. Do not merge. Do not append. Throw the work away. A late result that overwrites an earlier winner is data corruption, and the only safe move is to discard it.
Ini mengendalikan penyelesaian urutan terbalik dengan kemas. Percubaan A keluar dahulu tetapi kembali selepas tiga puluh saat. Percubaan B keluar kedua tetapi kembali selepas lima saat. Percubaan B memenangi compare-and-swap. Kemas kini Percubaan A tidak menyentuh sebarang baris. Sistem anda merekodkan perlumbaan tersebut, mengabaikan payload yang usang, dan meneruskan proses.
Uji Titik Kegagalan
Anda tidak akan menemui pepijat ini dalam ujian laluan senang (happy-path). Set ujian anda perlu menyasarkan titik-titik kelemahan tersebut.
- Simulasikan klik dua kali. Dua permintaan POST serentak dengan kunci idempotensi yang sama mesti mengembalikan ID tugasan yang serupa.
- Hantar kunci yang sama dengan input yang tidak sepadan. Jangkakan respons konflik. Sistem tidak boleh mengembalikan tugasan sedia ada secara senyap jika parameter berbeza.
- Cetuskan tamat masa (timeout). Sahkan bahawa tugasan berada dalam keadaan tidak diketahui, bukan keadaan gagal, dan sistem menyekat percubaan lanjut sehingga kekaburan tersebut selesai.
- Paksa dua percubaan untuk selesai dalam urutan terbalik. Sahkan bahawa percubaan kedua yang kembali akan kalah, walaupun yang pertama keluar adalah pembekal utama rasmi.
Ujian-ujian ini bukan sekadar kemewahan kes terpencil (edge-case). Ia adalah kontrak yang dibuat oleh API anda dengan sistem yang lain.
Sahkan Niat Pembekal Sebelum Anda Melakukan Failover
Jika anda menjalankan tetapan pelbagai pembekal, anda mungkin tergoda untuk menganggap model AI yang berbeza sebagai slot yang boleh ditukar ganti. Mereka berkongsi laluan kod yang sama, klien HTTP yang sama, dan skema JSON yang sama. Itu tidak bermakna mereka berkelakuan dengan cara yang sama.
Satu model mungkin mengalami halusinasi kunci peringkat teratas (top-level key). Model lain mungkin mengabaikan format arahan sistem (system prompt) anda. Pengesahan skema mengesan ralat sintaks, tetapi ia akan meluluskan respons yang tidak dapat ditafsirkan oleh logik perniagaan anda. Pembekal mungkin mengembalikan JSON yang sah tetapi hanya melakukan perkara yang salah dengan templat arahan anda.
Jalankan ujian khusus pembekal sebelum anda membenarkan penukaran model secara automatik. Sahkan bahawa model sandaran (fallback) benar-benar menghormati struktur output anda pada suhu (temperature) rendah. Sahkan bahawa arahan anda dipaparkan dengan betul melalui tokenizer pembekal tersebut. Uji perjalanan lengkap (round trip) dengan input sebenar. Failover automatik hanya selamat apabila anda telah membuktikan bahawa model sandaran berkongsi kontrak operasi yang sama.
Kekalkan Satu Tugasan Bagi Setiap Niat
Laluan sandaran adalah baik. Penggandaan sandaran yang tidak terkawal adalah satu pepijat. Setiap lapisan stack anda perlu menilai sama ada ia telah melihat tugasan yang tepat sebelum ini. Pengimbang beban (load balancer), pengendali API, pangkalan data, dan pekerja (worker) mesti semuanya menghormati identiti yang sama.
Bina sistem anda supaya percubaan semula (retries) dan sandaran muncul sebagai percubaan baharu di bawah satu tugasan yang stabil. Kunci tugasan tersebut dengan kunci idempotensi berasaskan pangkalan data. Lindungi peralihan tersebut. Biarkan percubaan bersaing. Biarkan hanya satu yang menang. Begitulah cara anda mengelakkan satu klik pengguna daripada bertukar menjadi hujung minggu untuk pembersihan data.