Our deploy ran composer update on a 2 GB server until the resolver ran out of memory. Raising the limit let the OOM killer take MySQL instead. Committing the lock file exposed a second bug: it had been resolved on PHP 8.4 for a PHP 8.2 server.
One CI run in eight failed somewhere in our checkout E2E suite, never in the same test twice. The causes were fixed sleeps, URL checks that raced navigation and text selectors that matched two buttons. Web-first assertions, role locators and traces fixed it.
Four engineers, one invoicing product, no QA team. This is the risk-based testing strategy we settled on: integration tests over mocks, contract tests at service edges, a handful of E2E happy paths and test data that does not rot.
Our CI compared LLM summaries against stored snapshots at temperature 0 and failed about one run in five. Temperature 0 is not a determinism guarantee on hosted models. We replaced string snapshots with property checks and a scored eval.
Semantic search in our docs assistant took 1.9 seconds and returned odd rankings. The query sorted by L2 distance while the HNSW index was built for cosine, so Postgres scanned every row. Matching the operator to the opclass brought it to 14 ms.
A VPC line item we had never paid attention to reached 2,400 USD a month. Since February 2024, AWS charges for every public IPv4 address, in use or idle. Our EKS nodes, idle Elastic IPs and load balancers in many subnets added up. Inventory and cleanup cut it by 70 percent.
A developer added user_id as a label on a request latency histogram to debug slow accounts. Active time series went from 1.2 million to 19 million, our hosted metrics bill quadrupled and the query layer slowed down. We found the culprit, dropped the label and set cardinality limits.
After we redeployed our MLflow tracking server on Kubernetes, every run older than the deploy showed metrics but no artifacts, and model loading failed. The server wrote artifacts to a local directory in the pod. We moved to proxied artifact storage on S3 and recovered what we could.
Two training runs with the same seed and data ended with validation accuracy 0.3 points apart, enough to flip a model comparison. Enabling deterministic algorithms then crashed with a CuBLAS error. Seeding every RNG, DataLoader workers, cuDNN flags and CUBLAS_WORKSPACE_CONFIG made runs repeatable.
Our rebalancing bot computed order sizes as fractions of a portfolio and about one order in five was rejected with Filter failure: LOT_SIZE or PRICE_FILTER. Quantities and prices must sit on each symbol's step and tick grid. Reading exchangeInfo filters and rounding with Decimal fixed it.
A trading system that cannot explain its balances to the cent will eventually make decisions on wrong numbers. This guide shows how we keep an internal double-entry ledger for trades, fees, funding and transfers, reconcile it with exchange balances, and handle the differences we find.
Our treasury rebalancing service swapped on a DEX and about a quarter of swaps reverted on volatile days with Too little received, wasting gas. Quotes were computed seconds before submission and transactions waited in the mempool. Fresh quotes, simulation and a deadline fixed it without loosening slippage.
After upgrading to Next.js 15 our dev console filled with sync dynamic API warnings, and one un-awaited auth helper quietly stopped redirecting logged-out users. How we migrated cookies(), headers() and params safely.
One slow partner API call made our whole FastAPI service stop answering, health checks included, until Kubernetes restarted the pods. A synchronous requests call and a time.sleep inside async def were blocking the event loop.
Our document tagger hit HTTP 429 constantly while sending only 40 requests a minute, far below the request limit. The token-per-minute limit was the real ceiling, inflated by an oversized max_tokens and minute-aligned bursts. How we read the headers and built a client-side limiter.
Our nightly embedding backfill died on the same few documents every night. A handful of chunks were far over the embedding model's token limit because the chunker split by characters and some files were minified JSON. Token-aware chunking and batching fixed it.
Our nightly product description job used the Batch API to save cost. One night the batch expired with 30 percent unfinished, and our importer joined results by line position, so descriptions landed on the wrong products. We now join by custom_id and handle every terminal status.
After upgrading to PyTorch 2.6, every checkpoint load in our evaluation pipeline failed with Weights only load failed. The default for weights_only in torch.load flipped to True, and our checkpoints contained a NumPy scalar. We fixed the checkpoints instead of turning the safety off.
Enabling dynamic batching in Triton Inference Server doubled throughput for our ranking model and pushed p99 latency from 38 to 85 ms, breaking our SLO. A large queue delay and preferred batch sizes made requests wait for batches that rarely filled at low traffic. Measured tuning fixed both.
A moving-average strategy showed strong backtest returns and nothing like it in paper trading. The backtest computed the signal from a candle's close and filled the order at that same close, a price nobody could trade at once the signal was known. Shifting signals to the next bar fixed the simulation.
After upgrading to React Native 0.79 with the New Architecture on by default, our in-house printer module came back null on iOS and crashed on Android. Bridge-only registration and direct bridge calls were the cause; here is the migration path.
Stock updates from supplier feeds started failing with Deadlock found when trying to get lock during peak hours. Concurrent upsert batches touched the same unique keys in different orders under InnoDB next-key locking. Sorting, smaller batches and a retry fixed it.
Requests decide where a pod is scheduled and how much CPU it is guaranteed; limits decide when it is throttled or killed. This guide explains QoS classes, CFS throttling and OOM kills, and shows how we size containers from real metrics with HPA and VPA.
Our docs assistant answered v4 questions with v3 settings, fluently and with citations. Fixed-size chunking split tables from their headings, chunks carried no version metadata and we had no retrieval eval. How we rebuilt chunking, added filters and hybrid search, and measured it.
After we set "type": "module" in package.json, jest.mock stopped replacing our mailer and tests started calling the real SMTP client. Static ESM imports are linked before any mock runs. Here is the mechanism and the fix with unstable_mockModule.