Apache Spark EOL: end-of-life dates for every Spark version
Apache Spark 3.4 reached end of life in October 2024 with 3.4.4 as its last release, and Spark 3.3 in December 2023. Spark 4.0's end-of-life date is 23 November 2026, 4.1's is 11 June 2027 and the 3.5 LTS line's is 30 November 2027. OSSeva patches Spark 3.4 and 3.3 after end of life. Dates checked against spark.apache.org and endoflife.date on 19 November 2026.
- Latest release
- 4.3.0
- Supported lines
- 5
- End-of-life lines
- 17
- Next end of life
- 4.0 · 23 Nov 2026
Spark 3.4 and 3.3 have ended
Move to 3.5 or 4.x, or keep 3.4 and 3.3 patched
Most jobs should move to a maintained line, but a move from 3.x to 4.x means Java 17, Scala 2.13 and ANSI SQL by default. When a library build, a vendor image or a release freeze holds a cluster on 3.4 or 3.3, OSSeva patches both. Other Spark lines are assessed case by case through a version coverage check. Critical fixes ship within 48 hours and High within 7 days.
Who patches Apache Spark after end of life
OSSeva provides Apache Spark extended support: patched, signed builds for Apache Spark 3.4 and 3.3 after upstream support ends, plus 24/7 operations on the Operate tier.
Apache Spark release timeline
All Apache Spark versions, release dates and end-of-life dates
| Version | Released | Support | Latest | OSSeva |
|---|---|---|---|---|
| 4.3 | 23 days ago(30 Oct 2026) | Ends in 1 year and 5 months(30 Apr 2028) | 4.3.0 | – |
| 4.2 | 4 months ago(14 Jul 2026) | Ends in 1 year and 2 months(11 Jan 2028) | 4.2.1 | – |
| 4.1 | 11 months ago(11 Dec 2025) | Ends in 7 months(11 Jun 2027) | 4.1.4 | – |
| 4.0 | 1 year and 6 months ago(19 May 2025) | Ends in 1 day(23 Nov 2026) | 4.0.5 | – |
| 3.5 (LTS) | 3 years and 2 months ago(9 Sept 2023) | Ends in 1 year(30 Nov 2027) | 3.5.9 | – |
| 3.4 | 3 years and 8 months ago(7 Apr 2023) | Ended 2 years and 1 month ago(21 Oct 2024) | 3.4.4 | Patched |
| 3.3 | 4 years and 5 months ago(9 Jun 2022) | Ended 2 years and 11 months ago(9 Dec 2023) | 3.3.4 | Patched |
| 3.2 | 5 years and 2 months ago(6 Oct 2021) | Ended 3 years and 7 months ago(9 Apr 2023) | 3.2.4 | – |
| 3.1 | 5 years and 9 months ago(22 Feb 2021) | Ended 4 years and 3 months ago(22 Aug 2022) | 3.1.3 | – |
| 3.0 | 6 years and 6 months ago(6 Jun 2020) | Ended 5 years ago(6 Dec 2021) | 3.0.3 | – |
| 2.4 (LTS) | 8 years and 1 month ago(29 Oct 2018) | Ended 5 years and 6 months ago(9 May 2021) | 2.4.8 | – |
| 2.3 | 8 years and 9 months ago(22 Feb 2018) | Ended 7 years and 3 months ago(25 Aug 2019) | 2.3.4 | – |
| 2.2 | 9 years and 5 months ago(30 Jun 2017) | Ended 7 years and 10 months ago(30 Jan 2019) | 2.2.3 | – |
| 2.1 | 9 years and 11 months ago(15 Dec 2016) | Ended 8 years and 5 months ago(26 Jun 2018) | 2.1.3 | – |
| 2.0 | 10 years and 4 months ago(19 Jul 2016) | Ended 9 years and 11 months ago(15 Dec 2016) | 2.0.2 | – |
| 1.6 | 10 years and 11 months ago(21 Dec 2015) | Ended 10 years and 4 months ago(19 Jul 2016) | 1.6.3 | – |
| 1.5 | 11 years and 2 months ago(8 Sept 2015) | Ended 10 years and 11 months ago(21 Dec 2015) | 1.5.2 | – |
| 1.4 | 11 years and 6 months ago(2 Jun 2015) | Ended 11 years and 2 months ago(8 Sept 2015) | 1.4.1 | – |
| 1.3 | 11 years and 9 months ago(5 Mar 2015) | Ended 11 years and 6 months ago(2 Jun 2015) | 1.3.1 | – |
| 1.2 | 11 years and 11 months ago(10 Dec 2014) | Ended 11 years and 8 months ago(5 Apr 2015) | 1.2.2 | – |
| 1.1 | 12 years and 3 months ago(3 Sept 2014) | Ended 11 years and 11 months ago(10 Dec 2014) | 1.1.1 | – |
| 1.0 | 12 years and 6 months ago(26 May 2014) | Ended 12 years and 3 months ago(3 Sept 2014) | 1.0.2 | – |
Lifecycle data from endoflife.date (MIT licence), refreshed 21 Nov 2026. Upstream policy: spark.apache.org. “Patched” marks end-of-life lines OSSeva ships signed, backported builds for.
Which Apache Spark versions are still supported?
The Spark versioning policy gives a 6-month window to major and feature releases and 18 months to an LTS release, and says critical security patches are backported to every branch that is still maintained. It states no end date for 3.3, 3.4, 4.0 or 4.1. The one date it does state is for 3.5.x, which has an extended LTS period to November 2027 with security fixes only. The end dates below for the other lines are from endoflife.date.
Release dates and latest releases are from spark.apache.org/news. endoflife.date shows 3.4.4, 3.3.4 and 4.1.0 a few days earlier than the project's news page.
Spark lines. Sources: spark.apache.org versioning policy and news, and endoflife.date, read on 19 November 2026.
| Line | First release | End of life | Latest release listed |
|---|---|---|---|
| 4.3 | 30 Oct 2026 | 30 Apr 2028 (endoflife.date) | 4.3.0 (30 Oct 2026) |
| 4.2 | 14 Jul 2026 | 11 Jan 2028 (endoflife.date) | 4.2.1 (27 Oct 2026) |
| 4.1 | 16 Dec 2025 | 11 Jun 2027 (endoflife.date) | 4.1.4 (27 Oct 2026) |
| 4.0 | 23 May 2025 | 23 Nov 2026 (endoflife.date) | 4.0.5 (27 Oct 2026) |
| 3.5 LTS | 13 Sep 2023 | 30 Nov 2027, security fixes only (project and endoflife.date) | 3.5.9 (16 Jul 2026) |
| 3.4 | 13 Apr 2023 | Oct 2024 (endoflife.date) | 3.4.4 (27 Oct 2024) |
| 3.3 | 16 Jun 2022 | Dec 2023 (endoflife.date) | 3.3.4 (16 Dec 2023) |
The 4.2 and 4.3 dates are 18 months after release, while the versioning policy describes a 6-month window for feature releases and says 4.2.x keeps its 18-month window. Treat them as endoflife.date's reading until the project states them. The policy also says 4.5 will be the 4.x LTS release.
What changes when Spark 3.x becomes Spark 4.x?
A cluster on 3.3, 3.4 or 3.5 cannot move to 4.0 by swapping the distribution. Spark 4.0 dropped JDK 8 and 11 and made JDK 17 the default, dropped Scala 2.12 and made 2.13 the default, dropped Python 3.8, turned ANSI SQL mode on by default and removed Mesos support. Every jar your jobs depend on that was built as _2.12 has to be replaced with a _2.13 build, which is usually the longest part of the move.
ANSI mode is the quiet change. The 4.0 release notes list SPARK-44444, using ANSI SQL mode by default, alongside fixes that return NULL on overflow when casting timestamps to integers and correct decimal multiplication and division. Queries that relied on silent overflow or lenient casts now fail or return different values, so a result comparison on production data belongs in the test plan, not just a clean build.
The requirements keep moving. The Spark 4.3.0 documentation says Spark runs on Java 17, 21 or 25 with Scala 2.13 and Python 3.11 or later, so a PySpark estate on Python 3.9 or 3.10 has to move its interpreter as well as Spark.
Platform requirements by line. Sources: Spark 3.5.9 and 4.3.0 documentation and the Spark 4.0.0 release notes, read on 19 November 2026.
| Area | Spark 3.5 | Spark 4.0 | Spark 4.3 |
|---|---|---|---|
| Java | 8, 11 or 17 | JDK 17 default; JDK 8 and 11 dropped | 17, 21 or 25 |
| Scala | 2.12 or 2.13 | 2.13 default; 2.12 dropped | 2.13 only |
| Python | 3.8 or later | 3.8 dropped | 3.11 or later |
| SQL semantics | ANSI mode off by default | ANSI mode on by default | Not re-checked |
| Mesos | Supported | Removed | Not re-checked |
The 4.0.0 release notes also list a minimum pyarrow of 11.0.0, a minimum NumPy of 1.21, Pandas upgraded to 2, removal of Hive support before 2.0.0, and SparkR as deprecated.
Which Spark vulnerabilities have no fix on 3.3?
CVE-2025-55039 is the case that shows what end of life costs. The Spark security page and the CVE record both say it affects Spark versions before 3.4.4, 3.5.2 and 4.0.0: when spark.network.crypto.enabled is true and spark.network.crypto.cipher is not set, RPC encryption falls back to AES in CTR mode, which has no authentication, so traffic between nodes can be modified. The CVE record's affected range starts at version 0, and the record was published on 15 October 2025.
3.3.4, released on 16 December 2023, was the last 3.3 release, so no 3.3 release carries the fix. A cluster on 3.3 has to set spark.network.crypto.cipher to AES/GCM/NoPadding or turn on SSL for RPC itself, or move lines. Both sources give the same two mitigations.
CVE-2025-55039. Sources: spark.apache.org/security.html and the cveawg.mitre.org record, read on 19 November 2026.
| Line | Status | Fixed in |
|---|---|---|
| 3.3 and earlier | Affected, line ended | No release on the line |
| 3.4 | Affected before 3.4.4 | 3.4.4 |
| 3.5 | Affected before 3.5.2 | 3.5.2 |
| 4.0 and later | Fixed at 4.0.0 | 4.0.0 |
It applies only when spark.network.crypto.enabled is true; the default is false. Both the Spark security page and the CVE record rate it moderate.
Who patches Apache Spark after end of life?
OSSeva patches Apache Spark 3.4 and 3.3 after end of life, and assesses other versions case by case through a [version coverage check](/technologies/apache-spark). Spark 3.5 has security-only maintenance to November 2027 from the project itself, so its date is the one to watch next after 4.0. This page does not compare other providers.
Running an end-of-life Apache Spark version?
OSSeva keeps end-of-life Apache Spark patched with signed builds and backported CVE fixes, documents it for your auditor, and can run it for you, so the upgrade happens on your schedule rather than the project’s.
Apache Spark extended supportIs Apache Spark 3.4 end of life?
Yes. endoflife.date records the end of life of Spark 3.4 as 21 October 2024, and the last release was 3.4.4, listed on spark.apache.org on 27 October 2024. The Spark versioning policy gives no separate end date for 3.4.
Is Apache Spark 3.3 end of life?
Yes. endoflife.date records December 2023 as the end of life of Spark 3.3, and 3.3.4 of 16 December 2023 was the last 3.3 release. That is why CVE-2025-55039, fixed in 3.4.4, 3.5.2 and 4.0.0, has no fix on 3.3.
When does Apache Spark 4.0 reach end of life?
Spark 4.0's end-of-life date is 23 November 2026, per endoflife.date. The spark.apache.org versioning policy does not state a date for 4.0. The latest 4.0 release is 4.0.5, from 27 October 2026.
Is Spark 3.5 an LTS release?
Yes. The versioning policy gives 3.5.x an extended LTS period that ends in November 2027, with security fixes only, and endoflife.date dates it 30 November 2027. The extension covers the main Spark repository, not Spark Connect for Swift, Rust or Go or the Kubernetes operator.
What is the latest Apache Spark version?
Spark 4.3.0, released on 30 October 2026 per spark.apache.org. 4.2.1, 4.1.4 and 4.0.5 came out three days earlier, on 27 October 2026.
Does Apache Spark 4 need a newer Java and Scala?
Yes. Spark 4.0 made JDK 17 the default and dropped JDK 8 and 11, and dropped Scala 2.12 for 2.13. The Spark 4.3.0 documentation lists Java 17, 21 or 25, Scala 2.13 and Python 3.11 or later.
Who provides Apache Spark support after end of life?
OSSeva patches Apache Spark 3.4 and 3.3 after end of life and assesses other versions case by case through a version coverage check. This page does not compare other providers.
When is the next Apache Spark end of life?
Apache Spark 4.0 reaches the end of support on 23 Nov 2026. After that date the project stops shipping security fixes for 4.0, and CVEs disclosed later stay unpatched upstream.
Which Apache Spark versions are still supported?
4.3 (until 30 Apr 2028), 4.2 (until 11 Jan 2028), 4.1 (until 11 Jun 2027), 4.0 (until 23 Nov 2026) and 3.5 (until 30 Nov 2027). Every other line has reached end of life, per the dates in the table above.
Can I get security patches for an end-of-life Apache Spark version?
Yes. OSSeva ships signed builds with backported CVE fixes for Apache Spark 3.4 and 3.3 after upstream support ends, on the Patch, Assure and Operate tiers, so you can stay on the version you run while you plan the upgrade.
Get a quote
Still running an end-of-life Apache Spark?
Tell us the versions and the estate size. We will tell you whether to upgrade or keep it patched, and what each costs.
Prefer to talk or read more first?