<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Nazelin Özalp on Medium]]></title>
        <description><![CDATA[Stories by Nazelin Özalp on Medium]]></description>
        <link>https://medium.com/@nazelin.ozalp?source=rss-a1ce67afa0a0------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*TWWCec9Q84NVI6wFoQunFQ@2x.jpeg</url>
            <title>Stories by Nazelin Özalp on Medium</title>
            <link>https://medium.com/@nazelin.ozalp?source=rss-a1ce67afa0a0------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 16:05:20 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@nazelin.ozalp/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Can a MongoDB Change Event Represent Two Different Moments in Time?]]></title>
            <link>https://medium.com/@nazelin.ozalp/can-a-mongodb-change-event-represent-two-different-moments-in-time-44ff04e5931b?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/44ff04e5931b</guid>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Mon, 02 Nov 2026 12:27:53 GMT</pubDate>
            <atom:updated>2026-11-02T12:27:53.780Z</atom:updated>
            <content:encoded><![CDATA[<h4><em>Understanding the gap between updateDescription, fullDocument, and what Change Streams actually promise</em></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CWR91j_Px3rJq6mNcnCOVQ.png" /><figcaption><em>The lookup for event 1 runs at t3, after both writes are committed, so the event carries state from two different moments.</em></figcaption></figure><p>We ran into this recently while chasing a problem in one of our Change Stream consumers. The events were arriving in order and nothing was missing, yet the data downstream did not match what had happened. The cause turned out to be a single option we had enabled without thinking twice about it.</p><p>Take an order document that starts as pending. One process sets it to paid; a few milliseconds later another sets it to shipped. You consume the first update event and see this:</p><ul><li>updateDescription says the status changed to paid.</li><li>fullDocument.status says shipped.</li></ul><p>Our first guess was a skipped event or an out-of-order stream. It was neither. This is how fullDocument: &quot;updateLookup&quot; is designed to work, and it means a single change event can describe one moment while carrying a document from a later one.</p><p>If you build audit trails, CDC pipelines or anything event-driven on Change Streams, this matters more than it first appears.</p><h3>1. What updateLookup actually returns</h3><p>By default, an update event carries only the delta:</p><pre>{<br>  operationType: &quot;update&quot;,<br>  documentKey: { _id: 1001 },<br>  updateDescription: {<br>    updatedFields: { status: &quot;paid&quot; },<br>    removedFields: [],<br>    truncatedArrays: []<br>  }<br>}</pre><p>Most consumers want the whole document, so they open the stream with fullDocument: &quot;updateLookup&quot;. MongoDB then <em>looks the document up</em> and attaches its most recent majority-committed version.</p><p>What tripped us up was the timing of that lookup. It does not happen when the write is applied, and it does not happen when your code processes the event. It happens on the server, when the event is prepared for a batch your cursor requests (the getMore). Anything majority-committed to that document in between is already in the result.</p><p>The Change Streams documentation says it directly: if other majority-committed operations modify the document between the update and the lookup, the returned document may differ significantly from the one at the time of the update.</p><h3>2. One event, two moments</h3><p>Take the order again: pending → paid → shipped. If the second update is majority-committed before the server looks up the document for the first event, the first event arrives like this:</p><pre>{<br>  operationType: &quot;update&quot;,<br>  documentKey: { _id: 1001 },<br>  updateDescription: {<br>    updatedFields: { status: &quot;paid&quot; },   // moment 1<br>    removedFields: [],<br>    truncatedArrays: []<br>  },<br>  fullDocument: {<br>    _id: 1001,<br>    customerId: 500,<br>    status: &quot;shipped&quot;,                 // moment 2<br>    totalAmount: 250<br>  }<br>}</pre><p>Nothing is lost or corrupted; the two fields just answer different questions. updateDescription tells you what <em>this</em> write changed. fullDocument tells you what the document looked like when the server went to fetch it.</p><h4>The extreme case: the document is gone</h4><p>The gap can be wider than a different status. Suppose the order is deleted right after it is paid:</p><pre>db.orders.updateOne({ _id: 1001 }, { $set: { status: &quot;paid&quot; } })<br>db.orders.deleteOne({ _id: 1001 })</pre><p>If the delete is majority-committed before the lookup, the update event looks like this:</p><pre>{<br>  operationType: &quot;update&quot;,<br>  documentKey: { _id: 1001 },<br>  updateDescription: {<br>    updatedFields: { status: &quot;paid&quot; },<br>    removedFields: [],<br>    truncatedArrays: []<br>  },<br>  fullDocument: null<br>}</pre><p>The update really happened, but the lookup found nothing. A consumer that assumes fullDocument is always present on update events will either crash on a null reference or silently skip the event. It rarely shows up in a happy-path test, which is exactly why it is worth handling explicitly.</p><h3>3. Reproducing it in a few lines</h3><p>You do not need two sessions or timing tricks. Because the lookup happens at getMore time, it is enough to open the stream, make both writes, and only then ask for the first event:</p><pre>db.orders.insertOne({ _id: 1001, status: &quot;pending&quot;, amount: 250 })<br>const stream = db.orders.watch([], { fullDocument: &quot;updateLookup&quot; })<br>db.orders.updateOne({ _id: 1001 }, { $set: { status: &quot;paid&quot; } })<br>db.orders.updateOne({ _id: 1001 }, { $set: { status: &quot;shipped&quot; } })<br>const e = stream.next()<br>printjson({ changed: e.updateDescription.updatedFields, doc: e.fullDocument })<br>// changed: { status: &quot;paid&quot; }, doc: { ..., status: &quot;shipped&quot; }</pre><p>The stream is opened before both writes, but its first batch is fetched only after both are committed, so the lookup sees shipped. Replace the second update with deleteOne and the same script gives you fullDocument: null.</p><p><em>Tested on MongoDB 8.0, single-node replica set.</em></p><p>This also means that adding a sleep in your consumer <em>after</em> the batch has been fetched changes nothing. The lookup has already happened.</p><h3>4. When you need the real moment: pre- and post-images</h3><p>MongoDB 6.0 added Change Stream pre- and post-images. Instead of a later lookup, the server stores the document as it was around each write and attaches it to that specific event.</p><p>Enable capture on the collection, then ask for the images:</p><pre>db.runCommand({<br>  collMod: &quot;orders&quot;,<br>  changeStreamPreAndPostImages: { enabled: true }<br>})<br>const stream = db.orders.watch([], {<br>  fullDocumentBeforeChange: &quot;required&quot;,<br>  fullDocument: &quot;required&quot;<br>})</pre><p>Now the first event carries exactly the transition it represents:</p><pre>{<br>  operationType: &quot;update&quot;,<br>  fullDocumentBeforeChange: { _id: 1001, status: &quot;pending&quot;, amount: 250 },<br>  fullDocument:             { _id: 1001, status: &quot;paid&quot;,    amount: 250 }<br>}</pre><p>The two options look alike but accept different values:</p><ul><li><strong>fullDocument</strong> accepts default, updateLookup, whenAvailable and required. updateLookup is the later lookup; whenAvailable and required return the stored post-image.</li><li><strong>fullDocumentBeforeChange</strong> accepts off, whenAvailable and required, and returns the stored pre-image.</li></ul><p>whenAvailable returns null if the image is missing. required raises an error instead, and the stream stops at that event until you deal with it. Choose required only if you have a plan for that failure, not just because it sounds safer.</p><p>updateLookup needs no collection setting. The image modes do.</p><h3>5. What pre- and post-images cost</h3><p>Images are stored in the internal config.system.preimages collection, and that has three kinds of cost.</p><p><strong>Writes.</strong> Every update or delete on an enabled collection also writes an image. On a write-heavy collection that is extra I/O and extra latency, not just extra disk.</p><p><strong>Storage and retention.</strong> By default, an image lives as long as its oplog entry: when the entry rolls off the oplog, the image goes too. You can cap it further:</p><pre>db.adminCommand({<br>  setClusterParameter: {<br>    changeStreamOptions: {<br>      preAndPostImages: { expireAfterSeconds: 86400 }<br>    }<br>  }<br>})</pre><p>With this set, an image is removed after 24 hours or when its oplog entry rolls off, whichever comes first. Images are also removed when the collection is dropped. A consumer that falls behind this window will find its images gone, so size the window from your real recovery requirement and monitor the size of config.system.preimages.</p><p><strong>Event size.</strong> A change event is still a BSON document with a 16 MB limit. With a pre-image and a post-image of a large document in the same event, you can hit it. MongoDB 7.0 added the $changeStreamSplitLargeEvent stage for exactly this case.</p><h3>6. Why correct ordering is not enough</h3><p>Change Streams deliver events in order. That is a guarantee about <em>events</em>, not about the documents attached to them.</p><p>Picture a notification service listening to the order stream. It sees event 1, reads updateDescription (paid), and sends &quot;Payment received&quot;. If it also reads fullDocument.status to pick a template, it sees shipped and sends &quot;Your order is on its way&quot; as well. Event 2 arrives with shipped and sends it again. The customer gets the shipping email twice, and the first one went out before your system had processed the shipping event.</p><p>In our case the symptom was quieter, just downstream data that did not match, but the mechanism was the same.</p><p>The same gap shows up in other forms:</p><ul><li>audit records that attach a later state to an earlier transition,</li></ul><p>projections that “jump ahead” and then replay the same state,</p><ul><li>consumers that crash or drop events on fullDocument: null,</li><li>replays that produce different results from the original run.</li></ul><p>Neither clusterTime nor the resume token closes this gap. clusterTime orders the event; it does not turn a later lookup into the state at that time. A resume token is a position in the stream, not a snapshot.</p><p>Resume tokens have their own trap: persist one only after the event’s side effect is done, or commit both atomically. Otherwise a crash between the two leads to duplicates or missed effects. Idempotent consumers make the whole problem much less painful.</p><h3>7. Picking the right tool</h3><ul><li><strong>Know which fields a write changed</strong> → updateDescription</li><li><strong>Get a convenient, recent copy of the document</strong> → updateLookup (and handle null)</li><li><strong>Know the exact before/after of each write</strong> → pre- and post-images</li><li><strong>Keep a long-term audit trail or replay history</strong> → pre/post-images <strong>plus</strong> your own durable event store</li></ul><p>Don’t skip the last one. Pre- and post-images are bounded by the oplog and your retention setting, so they are not an audit archive on their own. If you need history beyond that window, write the events you care about somewhere you control.</p><p>My take: most teams turn on updateLookup by default because it makes the consumer code simpler, and they never revisit that choice. That is fine for caches and search indexes, where only the latest state matters. But the moment a consumer makes a decision based on <em>what just happened</em>, such as sending a notification, writing an audit record or triggering a payment flow, I would stop relying on fullDocument from a lookup. Use updateDescription for the decision, or pay for pre- and post-images. Either way, make it a deliberate choice instead of a default.</p><h3>An event is not a snapshot</h3><p>updateDescription is what the write did. updateLookup is what the document looked like later, when the server fetched it. Pre- and post-images are what it looked like at the write.</p><p>An event can be perfectly ordered, successfully delivered, and still carry a document from a different moment. Before trusting fullDocument, check which of the three you actually asked for.</p><ul><li><a href="https://www.mongodb.com/docs/manual/changestreams/">MongoDB Change Streams - Database Manual - MongoDB Docs</a></li><li><a href="https://www.mongodb.com/docs/manual/reference/operator/aggregation/changestream/">$changeStream (aggregation stage) - Database Manual - MongoDB Docs</a></li><li><a href="https://github.com/mongodb/specifications/blob/master/source/change-streams/change-streams.md">specifications/source/change-streams/change-streams.md at master · mongodb/specifications</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=44ff04e5931b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Zero-Downtime Cluster Migration in Apache Cassandra]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/hepsiburadatech/zero-downtime-cluster-migration-in-apache-cassandra-6276cc47e2a2?source=rss-a1ce67afa0a0------2"><img src="https://cdn-images-1.medium.com/max/1445/1*0LeabHoH21GWD48p4OFO9w.jpeg" width="1445"></a></p><p class="medium-feed-snippet">Changing the operating system in distributed systems is often considered one of the highest-risk operations. Especially on critical data&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/hepsiburadatech/zero-downtime-cluster-migration-in-apache-cassandra-6276cc47e2a2?source=rss-a1ce67afa0a0------2">Continue reading on hepsiburadatech »</a></p></div>]]></description>
            <link>https://medium.com/hepsiburadatech/zero-downtime-cluster-migration-in-apache-cassandra-6276cc47e2a2?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/6276cc47e2a2</guid>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Tue, 31 Mar 2026 10:27:16 GMT</pubDate>
            <atom:updated>2026-03-31T10:27:16.817Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Ghosts in the Database: Understanding Dark Data]]></title>
            <link>https://medium.com/@nazelin.ozalp/ghosts-in-the-database-understanding-dark-data-5adc40439906?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/5adc40439906</guid>
            <category><![CDATA[dark-data]]></category>
            <category><![CDATA[database]]></category>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Mon, 18 Aug 2025 11:26:09 GMT</pubDate>
            <atom:updated>2025-08-18T11:26:09.112Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*zvb2_9Vwhdx_mQB4NRusqw.png" /><figcaption>Dark Data</figcaption></figure><p>Data is the most valuable asset of our time. Every company strives to collect more data, build larger databases, and perform more detailed analyses. But there’s a hidden truth we often overlook: the majority of the data we collect is never actually used. <em>This is where the concept of Dark Data comes in.<br>Dark Data is invisible yet persistent, unused yet costly, silent yet risky.</em></p><p>According to Gartner, more than half of the data companies collect is never utilized. Nearly every second table, every second log file, sits there unopened. And this invisible burden can become a serious problem.</p><h4>What is Dark Data?</h4><p>Simply put, Dark Data is data that isn’t used in business processes.<br> • Customer information collected years ago that no one queries,<br> • Leftover table copies from old campaigns,<br> • Log files kept “just in case,”<br> • Forgotten files in backup folders.</p><p>All of these are Dark Data. They sit quietly, yet every day they continue to generate costs and risks.</p><h4>The Hidden Cost</h4><p>The first impact of Dark Data is financial:<br> • <strong>Storage costs</strong>: Disks fill up, cloud bills rise.<br> • <strong>Backup costs</strong>: Even unused tables are backed up daily.<br> • <strong>Management costs</strong>: DBAs and data engineers spend time handling data that no one actually uses.</p><blockquote>Dark Data is a silent money pit.</blockquote><h4>The Bigger Threat: Security</h4><p>Cost is important, but the real risk of Dark Data is security. A table may go untouched for years, yet it may still contain personal data: national IDs, email addresses, credit card information.</p><p>If an attacker ever discovers that table, the most dangerous data is the data you never used.</p><p>Regulations like KVKK and GDPR are clear: keeping personal data you don’t use is a violation. Dark Data is not just a technical risk — it’s a legal one.</p><h4>Managing Dark Data</h4><p>Handling Dark Data takes courage. Business units often respond:</p><blockquote>“Shouldn’t we keep it? We might need it later.”</blockquote><p>But the time has come to move beyond the “we might need it” culture.<br> • <strong>Take inventory:</strong> Which data is actively used, and which is never queried?<br> • <strong>Archive:</strong> Move unused but potentially useful data to appropriate archives.<br> • <strong>Delete:</strong> Remove data that serves no purpose.<br> • <strong>Automate:</strong> Set up automated processes to track usage regularly.</p><p>These steps lighten databases and reduce security risks.</p><h4>Turning Dark Data into Value</h4><p>Not all Dark Data is waste. With the right approach, even hidden data can create opportunities:<br> • Old log files can help predict future errors.<br> • Customer behavior data can feed AI models.<br> • IoT data can reveal new trends and business opportunities.</p><p>The point isn’t just to store data, it’s to make it meaningful.</p><p><strong>Dark Data is the hidden weight of your databases</strong>:<br> • It silently increases costs,<br> • Creates security risks,<br> • Complicates management processes.</p><p>But when managed properly, Dark Data can also become a valuable asset.</p><blockquote>Ultimately, the question is:<br>Do we manage the data, or does the data manage us?</blockquote><p>💬 Do you have tables in your systems that are never queried? Is it harder to delete them or to keep them? Let’s discuss in the comments 👇</p><p><a href="https://www.dataversity.net/dark-data-explained/">Dark Data Explained - DATAVERSITY</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5adc40439906" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[CPU 93%, Focus 37%: When Logs Reflect the Engineer]]></title>
            <link>https://medium.com/@nazelin.ozalp/cpu-93-focus-37-when-logs-reflect-the-engineer-49f8fe35ae62?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/49f8fe35ae62</guid>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Tue, 17 Jun 2025 09:43:55 GMT</pubDate>
            <atom:updated>2025-06-17T09:43:55.111Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Some mornings, I start reading logs like I’m flipping through a journal.<em><br></em>And more often than not, I don’t just find the system there…<em><br></em>I find myself.</blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/44/0*Z1pb_rLg2xh-aw1r.png" /></figure><p>I Saw Myself in a Log Line<br>Last week, one morning, my phone buzzed.<br> I hadn’t even sipped my coffee yet when this alert popped up:</p><pre>[WARNING] CPU utilization exceeded 93% on node-03.</pre><p>Classic stuff. My first reflex was to check the query, the index, the cache.<br> Then I paused.<br>The week before, code reviews had piled up. I’d stayed up past 1 AM.<br> I’d skipped optimization, telling myself,<br><em>“I’ll fix it later.”</em><br>That moment it hit me:<br> <strong>This wasn’t just a system issue. It was my fatigue. My procrastination. My drifting focus.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/44/0*LJh8d_9_jD07niRz.png" /></figure><p><strong>Monitoring = A Practice of Awareness<br></strong>Monitoring tools always remind me of meditation.<br> Grafana, Kibana, ELK, Prometheus — they’re not just dashboards.<br> They’re whispering something if you know how to listen.<br>Log levels start to feel like emotional states:<br><strong>Log LevelWhat It Says About You</strong>ERRORThe moments you panicWARNINGThe gut feelings you suppressINFOYour daily autopilotDEBUGWhat surfaces when you look deeper</p><blockquote><em>If </em><em>DEBUG is turned off, you won’t see the details.<br> Same goes for yourself — if you avoid introspection, you only fight symptoms.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/44/0*x7x_1f85aIcR47qW.png" /></figure><p><strong>Alert Fatigue vs. Burnout<br></strong>Alert fatigue is real.<br> We monitor 2,000+ metrics across 20+ systems. Every day, something triggers.<br>But after a while, you stop hearing what actually matters.<br>Just like in life:</p><ul><li>Run to every alert → burnout</li><li>Don’t filter → overwhelmed</li><li>No prioritization → chaos</li></ul><blockquote><em>Monitoring teaches you something critical:<br> Just like metrics, </em><strong><em>emotions</em></strong><em> need filtering, thresholds, and alert logic too.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/44/0*CbCOgozqHSPhYYgZ.png" /></figure><p><strong>Dashboards as Mindfulness<br></strong>These days, when I check a dashboard in the morning, I don’t just see system health.<br> I see my team’s mental state.</p><ul><li>What time was the last job run?</li><li>Who retried the same query five times?</li><li>Who manually killed that batch process?</li></ul><p>Logs carry more than data.<br> <strong>They carry human fingerprints.</strong><br>Looking at a dashboard can feel like a mindfulness ritual:<br> You say good morning to the system — and to yourself:</p><blockquote><em>“How am </em>I<em> doing today?”</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/44/0*M8agtX7HuBjPPyhd.png" /></figure><p><strong>EXPLAIN ANALYZE: Not Just for PostgreSQL<br></strong>In databases, you use EXPLAIN ANALYZE to understand what your query <em>really</em> does.<br>But in life, you’ve got your own queries.<br>Ask yourself:</p><ul><li>“Why did I react that way?”</li><li>“Why do I keep repeating this mistake?”</li><li>“Why do I keep putting this task off?”</li></ul><blockquote><em>Learning to run </em><em>EXPLAIN ANALYZE on yourself might be the most underrated soft skill in engineering.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/44/0*cBQfhY6akn3UtBo8.png" /></figure><p><strong>Final Thought:<br></strong>If You Can’t Observe Yourself, You Can’t Observe Your Systems<br>Logs may not literally describe you.<br> But how you <strong>read</strong> them says a lot about who you are.<br>Monitoring systems aren’t just about alerts and metrics.<br> They reflect your habits, your priorities, your fears, your blind spots.</p><blockquote><em>Every </em><em>WARNING might be an inner signal.<br> Every </em><em>ERROR could echo a past decision.<br> And every </em><em>DEBUG... might point to your next step in growth.</em></blockquote><blockquote>Before you close your eyes tonight, read a log.<em><br></em>You might not just find a system issue — <em><br></em>you might find something about yourself.</blockquote><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=49f8fe35ae62" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Seamless PostgreSQL Migration from VMware to OpenStack: A Step-by-Step Guide]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@nazelin.ozalp/seamless-postgresql-migration-from-vmware-to-openstack-a-step-by-step-guide-147efcbbb148?source=rss-a1ce67afa0a0------2"><img src="https://cdn-images-1.medium.com/max/686/0*M37fXLkiFB5sslaA.jpeg" width="686"></a></p><p class="medium-feed-snippet">Following the successful migration of our MongoDB environment to OpenStack, we are now applying the same process for PostgreSQL. In this&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@nazelin.ozalp/seamless-postgresql-migration-from-vmware-to-openstack-a-step-by-step-guide-147efcbbb148?source=rss-a1ce67afa0a0------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@nazelin.ozalp/seamless-postgresql-migration-from-vmware-to-openstack-a-step-by-step-guide-147efcbbb148?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/147efcbbb148</guid>
            <category><![CDATA[postgresql]]></category>
            <category><![CDATA[openstack]]></category>
            <category><![CDATA[vmware]]></category>
            <category><![CDATA[postgres]]></category>
            <category><![CDATA[replication]]></category>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Wed, 22 Jan 2025 13:46:24 GMT</pubDate>
            <atom:updated>2025-01-22T13:46:24.776Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Achieving Zero Downtime: Migrating Live MongoDB Databases from VMware to OpenStack]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@nazelin.ozalp/achieving-zero-downtime-migrating-live-mongodb-databases-from-vmware-to-openstack-0db441b78913?source=rss-a1ce67afa0a0------2"><img src="https://cdn-images-1.medium.com/max/686/1*Ezpl1ShJ3Q64EAUKnXdsOw.jpeg" width="686"></a></p><p class="medium-feed-snippet">Transitioning from VMware to OpenStack is no small feat, especially for a high-traffic e-commerce platform like Hepsiburada, where orders&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@nazelin.ozalp/achieving-zero-downtime-migrating-live-mongodb-databases-from-vmware-to-openstack-0db441b78913?source=rss-a1ce67afa0a0------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@nazelin.ozalp/achieving-zero-downtime-migrating-live-mongodb-databases-from-vmware-to-openstack-0db441b78913?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/0db441b78913</guid>
            <category><![CDATA[openstack]]></category>
            <category><![CDATA[vmware]]></category>
            <category><![CDATA[postgresql]]></category>
            <category><![CDATA[mongodb]]></category>
            <category><![CDATA[mysql]]></category>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Mon, 12 Aug 2024 12:19:36 GMT</pubDate>
            <atom:updated>2024-08-12T12:19:36.969Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Seamless Migration: Transitioning from Google Cloud SQL for MySQL to On-Premise Servers with Zero…]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@nazelin.ozalp/seamless-migration-transitioning-from-google-cloud-sql-for-mysql-to-on-premise-servers-with-zero-28a254a5d8d8?source=rss-a1ce67afa0a0------2"><img src="https://cdn-images-1.medium.com/max/1274/1*38EHGVE3byfPms5VYYJDxw.png" width="1274"></a></p><p class="medium-feed-snippet">We wanted to share the steps we followed to migrate our Google Cloud SQL environments to on-prem servers without any downtime. The&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@nazelin.ozalp/seamless-migration-transitioning-from-google-cloud-sql-for-mysql-to-on-premise-servers-with-zero-28a254a5d8d8?source=rss-a1ce67afa0a0------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@nazelin.ozalp/seamless-migration-transitioning-from-google-cloud-sql-for-mysql-to-on-premise-servers-with-zero-28a254a5d8d8?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/28a254a5d8d8</guid>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Fri, 19 Jan 2024 05:41:01 GMT</pubDate>
            <atom:updated>2024-01-19T05:45:54.731Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Real-Time Monitoring and Swift Alerting for COLLSCAN, SLOW Queries in MongoDB with ELK Stack]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@nazelin.ozalp/real-time-monitoring-and-swift-alerting-for-collscan-slow-queries-in-mongodb-with-elk-stack-70aa9dc59141?source=rss-a1ce67afa0a0------2"><img src="https://cdn-images-1.medium.com/max/600/1*YrZTMq0VgLxarNUNWvY8Dg.png" width="600"></a></p><p class="medium-feed-snippet">MongoDB Performance Monitoring with ELK Stack</p><p class="medium-feed-link"><a href="https://medium.com/@nazelin.ozalp/real-time-monitoring-and-swift-alerting-for-collscan-slow-queries-in-mongodb-with-elk-stack-70aa9dc59141?source=rss-a1ce67afa0a0------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@nazelin.ozalp/real-time-monitoring-and-swift-alerting-for-collscan-slow-queries-in-mongodb-with-elk-stack-70aa9dc59141?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/70aa9dc59141</guid>
            <category><![CDATA[mongodb]]></category>
            <category><![CDATA[elk]]></category>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Tue, 31 Oct 2023 08:59:16 GMT</pubDate>
            <atom:updated>2023-11-01T11:21:08.344Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[Creating a Highly Available PostgreSQL Cluster with Patroni, Etcd: A Guide to Fault Tolerance and…]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://medium.com/@nazelin.ozalp/creating-a-highly-available-postgresql-cluster-with-patroni-etcd-a-guide-to-fault-tolerance-and-c8f865e3447?source=rss-a1ce67afa0a0------2"><img src="https://cdn-images-1.medium.com/max/600/1*hq4liVrsRo5vGcxAeT7ZNQ.png" width="600"></a></p><p class="medium-feed-snippet">In this article, I describe how I automated the installation of a highly available PostgreSQL cluster using Patroni, etcd, by building an&#x2026;</p><p class="medium-feed-link"><a href="https://medium.com/@nazelin.ozalp/creating-a-highly-available-postgresql-cluster-with-patroni-etcd-a-guide-to-fault-tolerance-and-c8f865e3447?source=rss-a1ce67afa0a0------2">Continue reading on Medium »</a></p></div>]]></description>
            <link>https://medium.com/@nazelin.ozalp/creating-a-highly-available-postgresql-cluster-with-patroni-etcd-a-guide-to-fault-tolerance-and-c8f865e3447?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8f865e3447</guid>
            <category><![CDATA[patroni]]></category>
            <category><![CDATA[postgres]]></category>
            <category><![CDATA[etcd]]></category>
            <category><![CDATA[ha]]></category>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Thu, 23 Feb 2023 08:15:36 GMT</pubDate>
            <atom:updated>2023-02-23T08:18:53.320Z</atom:updated>
        </item>
        <item>
            <title><![CDATA[ClickHouse Replication Using Keeper]]></title>
            <description><![CDATA[<div class="medium-feed-item"><p class="medium-feed-image"><a href="https://blog.devgenius.io/clickhouse-replication-using-keeper-219ca35821d?source=rss-a1ce67afa0a0------2"><img src="https://cdn-images-1.medium.com/max/2600/0*M0qH3JIKWRJA6oia.png" width="2600"></a></p><p class="medium-feed-snippet">ClickHouse has some differences from the standard SQL developed for Yandex, which is column-based, distributed, parallel processing&#x2026;</p><p class="medium-feed-link"><a href="https://blog.devgenius.io/clickhouse-replication-using-keeper-219ca35821d?source=rss-a1ce67afa0a0------2">Continue reading on Dev Genius »</a></p></div>]]></description>
            <link>https://blog.devgenius.io/clickhouse-replication-using-keeper-219ca35821d?source=rss-a1ce67afa0a0------2</link>
            <guid isPermaLink="false">https://medium.com/p/219ca35821d</guid>
            <category><![CDATA[clickhouse]]></category>
            <category><![CDATA[clickhousedb]]></category>
            <dc:creator><![CDATA[Nazelin Özalp]]></dc:creator>
            <pubDate>Mon, 28 Nov 2022 08:22:48 GMT</pubDate>
            <atom:updated>2022-12-06T12:19:51.592Z</atom:updated>
        </item>
    </channel>
</rss>