{"id":12818,"date":"2026-11-16T21:55:00","date_gmt":"2026-11-16T21:55:00","guid":{"rendered":"https:\/\/www.wakilsahab.in\/news\/?p=12818"},"modified":"2026-11-19T20:55:00","modified_gmt":"2026-11-19T20:55:00","slug":"serverlessops-daily-digest-steps","status":"publish","type":"post","link":"https:\/\/www.wakilsahab.in\/news\/serverlessops-daily-digest-steps\/","title":{"rendered":"ServerlessOps Steps for a Daily Digest That Never Posts Twice"},"content":{"rendered":"<p>A daily digest looks simple from the reader&#8217;s side: one page, every morning, the same format. Behind it sits a chain of small jobs. One fetches the day&#8217;s items, another trims and tags them, a third assembles the page, and a fourth pushes it live. When those jobs run as cloud functions triggered by events, you are doing ServerlessOps whether or not anyone on the desk uses the term. This guide goes through the steps in order, written for a small editorial-technology team that wants the digest to appear once, on time, with nothing missing.<\/p>\n<p>It is general operations advice. It says nothing about the legal developments themselves, and nothing here is legal guidance. The subject is how to keep the machinery that carries a digest from misbehaving. If you read daily roundups such as <a href=\"https:\/\/www.wakilsahab.in\/news\/top-20-indian-legal-developments-2026-june-19\/\">the newest daily legal-news roundup from mid-June<\/a> or <a href=\"https:\/\/www.wakilsahab.in\/news\/top-20-indian-legal-developments-2026-june-18\/\">its predecessor from the previous morning<\/a>, you are looking at exactly the kind of output where a missing or duplicated item is noticed fast.<\/p>\n<h2>Step 1: List every trigger before you touch any code<\/h2>\n<p>Start with a plain inventory. Write down each place where a function is started: a timer that fires at a set hour, a queue message, a file landing in storage, a webhook from a content tool. For each one, note whether the caller waits for the answer (synchronous) or hands the work off and moves on (asynchronous). That single distinction decides what happens when something fails, because the two models retry in very different ways.<\/p>\n<p>A timer-fired fetch job is asynchronous, so the platform may retry it on its own. A request from an editor pressing a button is synchronous, so the retry decision belongs to whoever pressed it. The Ops Atlas entry on this discipline lists the invocation model as the first thing to pin down, and in practice most desk incidents trace back to a trigger nobody had written down.<\/p>\n<h2>Step 2: Decide what &#8220;done&#8221; means in business terms<\/h2>\n<p>Platform dashboards report whether a function returned without throwing. That is not the same as the digest having been published. For a news desk, the real success event is &#8220;today&#8217;s page exists, once, with the expected number of items.&#8221; Write that sentence on the wall of your runbook.<\/p>\n<p>From this definition you can derive a business success rate: the share of mornings on which the page went live correctly. It is a more honest number than the function error count, because a function can finish happily while quietly skipping half the items.<\/p>\n<h2>Step 3: Give every item a key that belongs to the item<\/h2>\n<p>The most common mistake in a digest pipeline is using the message identifier as the duplicate guard. Message identifiers change when a stuck message is redriven from a failure queue; the replayed message gets a fresh one, so a guard keyed on it waves the duplicate straight through. Use a key that comes from the thing itself instead. For a news item that could be the source address combined with the publication date, or an internal story number.<\/p>\n<p>Store that key in a small table with an expiry, and check it before the function does anything that cannot be undone, such as publishing, emailing subscribers or posting to a social channel. If the key is already there, the function exits quietly. Pass the same key along to any downstream service that supports it, so the protection extends past your own code.<\/p>\n<p>Longer than the longest outage you want to survive, and no longer than the downstream service remembers. If your guard expires after an hour but a stuck queue is replayed after a day, the duplicate sails through. The Atlas notes that one popular helper library defaults to a one-hour expiry while its own page recommends fifteen days for its worked example, which shows how far apart the defaults and the sensible values can be. Pick the number deliberately.<\/p>\n<h2>Step 4: Put a failure destination on every asynchronous path<\/h2>\n<p>When an asynchronous function keeps failing, the platform eventually gives up. What happens next is up to you. If no dead-letter queue or failure destination is configured, the event simply disappears, and the first sign is a reader asking why yesterday&#8217;s item never appeared.<\/p>\n<p>For each asynchronous trigger from Step 1, attach a place for failed events to land. Then treat that place as a real piece of the system, with three properties:<\/p>\n<ul>\n<li><strong>An owner.<\/strong> A named person or rota, not &#8220;the team.&#8221;<\/li>\n<li><strong>An alarm.<\/strong> It should fire when anything arrives in the queue and again when the oldest item gets stale.<\/li>\n<li><strong>A retention period longer than the source.<\/strong> The Atlas gives an illustration of a failure queue kept fourteen days against a source queue that holds four, so an item cannot age out of the safety net before anyone looks.<\/li>\n<\/ul>\n<p>A failure queue nobody reads is just a slower way of losing items.<\/p>\n<h2>Step 5: Settle the timeouts from the outside in<\/h2>\n<p>Every call in the chain needs a time limit, and the limits must nest. The call your function makes to a downstream service should give up before the function itself does; otherwise the function is killed mid-call and you learn nothing about what the downstream service did. If a queue feeds the function, its visibility window needs to be comfortably longer than the function&#8217;s timeout. The Atlas suggests at least six times as long, which leaves room for a retry or two without the message reappearing while the first attempt is still running.<\/p>\n<p>For a digest that runs once in the morning, generous limits are cheap. A slow source feed at half past five is far less harmful than a function that is cut off without finishing and leaves a half-built page.<\/p>\n<h2>Step 6: Handle batches item by item<\/h2>\n<p>If your fetch step pulls twenty items in one batch and the seventeenth is malformed, an all-or-nothing response makes the platform retry all twenty, including the sixteen already processed. Return a partial response instead, listing only the failed item, so the rest are not repeated. For stream-style sources, also set a finite retry count and a maximum record age; the Atlas points out that the defaults for stream mappings can be unlimited, which turns one bad record into a blocked lane that holds up everything behind it.<\/p>\n<p>It also warns about a subtle alarm trap. Item-level failures reported through a partial response are not counted as function errors, so an alert wired only to the error metric stays silent while items pile up. Add alarms on backlog age and on failure-queue inflow so these cases are caught.<\/p>\n<h2>Step 7: Cap concurrency so a spike cannot hurt the neighbours<\/h2>\n<p>A busy news day can fan out into hundreds of simultaneous function runs, each opening a connection to the same content database or search index. Without a ceiling, the functions scale happily and the database falls over. Set a reserved or maximum concurrency on the functions that touch shared systems, sized from the downstream limit, not from the platform&#8217;s own ceiling. A practical rule from the Atlas suggests that expected peak load should stay under about seven-tenths of the ceiling you hold, and it adds that a reserved allocation needs enough unreserved capacity left in the account for everything else.<\/p>\n<h2>Step 8: Watch four numbers, not forty<\/h2>\n<p>Dashboards fill up quickly and then nobody reads them. For a digest, four signals cover most of the ground:<\/p>\n<ol>\n<li>Did today&#8217;s page publish exactly once? That is the business success check from Step 2.<\/li>\n<li>How many items are waiting in the failure queue, and how old is the oldest?<\/li>\n<li>How old is the oldest unprocessed event in the main flow (backlog age)?<\/li>\n<li>How close did concurrency come to its cap?<\/li>\n<\/ol>\n<p>Everything else, including cold-start share and cost per run, can be reviewed monthly. If you want a fuller catalogue, <a href=\"https:\/\/www.xopsschool.com\/serverlessops\/\">the Ops Atlas page covering this practice<\/a> lists the longer set and shows a worked example of turning counts into ratios.<\/p>\n<h2>Step 9: Rehearse the duplicate storm<\/h2>\n<p>The test that matters most is also the cheapest. Take a staging copy of the pipeline and deliver the same day&#8217;s items three times in a row, once normally, once as a replay and once as a deliberate overlap with a retry. The result should be one page and zero repeated notifications. If the digest posts twice, you have found the gap before a reader did.<\/p>\n<p>Repeat the exercise after any change to the publishing step. Two minutes of rehearsal is cheaper than a correction notice.<\/p>\n<h2>Step 10: Write the one-page runbook<\/h2>\n<p>Keep it to a single sheet. It should answer, in order: what are the triggers, who owns each failure queue, how do I replay safely, how long are the duplicate keys valid, and who do I call when the alarm rings at a bad hour. Link it from the editorial tools people already open every day. A runbook is easiest to trust when it mirrors the output, and a fixed daily format like <a href=\"https:\/\/www.wakilsahab.in\/news\/top-20-indian-legal-developments-2026-june-17\/\">the issue dated two days earlier<\/a> is easy to check against the day before.<\/p>\n<h2>Keep the upkeep small<\/h2>\n<p>Every Monday, someone spends fifteen minutes on the failure queue and writes one line per item. Every month, compare cost per run with the last one. Every quarter, repeat the duplicate rehearsal. Reliable publishing is mostly the absence of surprises.<\/p>\n<p>The sequence is the takeaway. Inventory triggers, define success in business terms, key by the item, catch failures, nest timeouts, report per item, cap concurrency, watch four numbers, rehearse, document. Do them in that order, and the digest keeps arriving once each morning without anyone staying up to babysit it.<\/p>\n<div class=\"fb-background-color\" style=\"background-color:#ffffff\"><div class=\"fb-comments\" style=\"width:100%\" data-href=\"https:\/\/www.wakilsahab.in\/news\/serverlessops-daily-digest-steps\/\" data-numposts=\"5\" data-lazy=\"true\" data-colorscheme=\"light\" data-order-by=\"social\" data-mobile=\"true\"><\/div><\/div>","protected":false},"excerpt":{"rendered":"<p>A practical walk-through for teams that publish a daily legal-news digest from event-driven functions: retries, dead-letter queues, and duplicate-proof publishing.<\/p>\n","protected":false},"author":6,"featured_media":12819,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[10013],"tags":[],"class_list":["post-12818","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-legal-tech"],"_links":{"self":[{"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/posts\/12818","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/users\/6"}],"replies":[{"embeddable":true,"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/comments?post=12818"}],"version-history":[{"count":1,"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/posts\/12818\/revisions"}],"predecessor-version":[{"id":12942,"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/posts\/12818\/revisions\/12942"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/media\/12819"}],"wp:attachment":[{"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/media?parent=12818"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/categories?post=12818"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.wakilsahab.in\/news\/wp-json\/wp\/v2\/tags?post=12818"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}