MySQL Master-Slave Replication for Zero-Downtime Failover on Ubuntu

We configure MySQL master–slave replication on Ubuntu so your apps keep serving reads—and you can promote a healthy slave—with a planned path to zero-downtime failover when the primary must leave the write path.

What You'll Get

• Hardened master configuration - Binlog format, server-id, GTID mode, and retention sized for your RPO without flooding disk.
• Replica bootstrap & sync - Clean dump or clone path, CHANGE REPLICATION SOURCE setup, and verified catch-up to near-zero lag.
• Failover readiness pack - Promotion checklist, read_only/super_read_only handling, and application DSN cutover notes.
• Monitoring hooks - Seconds_Behind_Source, IO/SQL thread health, and alert-friendly status queries you can drop into existing tools.
• Plain-English runbook - Who promotes, in what order, and how to re-attach the old master as a new slave after recovery.
• Optional dual-tier support - Technical deep-dive for your ops lead plus calm, jargon-light walkthroughs for non-technical stakeholders.

Serving Zero-Downtime Failover & surrounds

Zero-downtime failover is the reason this page exists: businesses that treat database write loss as a commercial incident, not a maintenance footnote. We work remote-first with teams whose peaks look like CBD lunch-order platforms, industrial warehouse WMS cutovers, and multi-clinic booking windows—not generic “high availability” brochure copy.

• Retail and café-strip stacks that need read scale on evenings and weekend markets without dual-write chaos
• Freight-season and warehouse-precinct systems where a stuck replica during dispatch windows costs real dollars
• Fully remote delivery with scheduled change windows; on-site only if your policy requires a supervised console handoff

How We Work

1. Step 1: Reach Out - Share Ubuntu versions, MySQL major, data size, current backup method, and whether apps already use separate read/write endpoints—we listen before proposing topology.
2. Step 2: Tailored Plan - Fixed-scope quote for Basic replica stand-up through Premium multi-replica failover drills; clear support path if your ops team needs a guided rehearsal.
3. Step 3: We Deliver - Remote configuration on your hosts or approved jump box: master prep, replica sync, privilege hardening, and controlled lag validation under realistic load samples.
4. Step 4: Confirm & Follow-up - Handoff with commands that worked in your environment, optional monitoring follow-up, and a short debrief so the next person on-call is not guessing.

Common Issues & How to Fix Them

These are the failure modes we see repeatedly on Ubuntu MySQL replicas aimed at zero-downtime failover—not generic “check the docs” tips.

Replica IO thread stops after a brief network blip on cloud or VPN paths

Symptoms: Slave_IO_Running = No, Last_IO_Error mentions connection reset or EOF, while the SQL thread is still fine—common on flaky site-to-site links and under-provisioned security groups.

1. Step 1: On the replica run SHOW REPLICA STATUS\G (or SHOW SLAVE STATUS) and note Master_Log_File / Read_Master_Log_Pos plus the exact IO error timestamp.
2. Step 2: Confirm the master is reachable on 3306 from the replica host (nc or mysql client with the replication user), then START REPLICA IO_THREAD; if errors mention SSL or auth, fix user REQUIRE SSL and host grants before looping restarts.
3. Step 3: Verify Slave_IO_Running = Yes and Seconds_Behind_Source trends down; if it reconnects then dies again, raise net timeouts and check intermediate firewall idle policies—not only MySQL.

GTID auto-positioning fails after restoring a backup taken without gtid_purged set

You see ERROR 1236 or “purged binary logs” style complaints when the replica tries to auto-position; teams feel this after “quick” mysqldump restores that skipped GTID metadata.

1. Step 1: On the master run SHOW VARIABLES LIKE 'gtid_executed'; and compare with the replica’s Retrieved_Gtid_Set / Executed_Gtid_Set after the failed START REPLICA.
2. Step 2: Restore using a dump that includes --set-gtid-purged=ON (or set gtid_purged from a consistent snapshot), RESET REPLICA ALL carefully on a disposable replica only, then reconfigure CHANGE REPLICATION SOURCE... SOURCE_AUTO_POSITION=1.
3. Step 3: Confirm both threads run and a test row inserted on the master appears on the replica with matching GTID history—never promote until Executed_Gtid_Set is coherent.

Failover “works” but apps keep writing the demoted master because DNS/DSN lag

MySQL promotion succeeded, yet zero-downtime goals fail because connection pools still point at the old writer; clinics and multi-store POS stacks feel phantom inventory and double bookings.

1. Step 1: After STOP REPLICA and SET GLOBAL read_only=0 / super_read_only=0 on the new primary, immediately set the old master read_only=1 and super_read_only=1 before touching DNS.
2. Step 2: Cut application write DSN (or proxy primary endpoint) first; recycle app pools or ProxySQL/HAProxy writer backends so idle connections cannot reopen to the demoted host.
3. Step 3: Verify with a canary UPDATE on the new primary and confirm the old master rejects writes; only then re-point the old master as a replica of the new primary.

When DIY is not enough (urgent, unsafe, recurring, or burning time), book Fixwebnode for direct professional support—no freelancers, bidding, or marketplace noise.

Why Choose Fixwebnode

We are a direct provider for Ubuntu MySQL replication and failover readiness—not a bid board. You get specialist configuration plus human-clear runbooks so on-call staff can execute promotion without decoding tribal knowledge at 2 a.m.

• ✓ Hands-on GTID, binlog, and lag triage across production Ubuntu LTS fleets
• ✓ Fixed-scope packages with written promotion order—not open-ended hourly mystery
• ✓ Dual-tier delivery: deep technical work for engineers, plain-language briefings for owners and clinic managers

Tools & Technologies

MySQL 8.0 / 5.7 on Ubuntu LTS, GTID replication, mysqldump and mydumper/myloader where appropriate, mysqlbinlog analysis, Percona Toolkit (pt-table-checksum / pt-table-sync when scoped), ProxySQL or HAProxy writer/reader split guidance, systemd unit hardening, UFW/security-group port discipline, and status automation via simple shell or existing Prometheus exporters.

Perfect For

SaaS and internal platforms that need read scale today and a rehearsed writer promotion path for true zero-downtime failover goals; clinics, education portals, and small ops teams who want enterprise-grade replication without a marketplace of unknown hands. Ideal when one primary still owns writes and you need a trustworthy slave—not a half-finished dual-master experiment.

Next step: outline your Ubuntu/MySQL versions and change window, then reach Fixwebnode support or chat with us to lock a Basic, Standard, or Premium scope.