Skip to main content

Fixing Hive “Too Many Counters” MapReduce Failures

By Anonymous
When Hive queries use many operators, MapReduce can hit its default counter limit and fail with a “Too many counters” exception. This short note explains why it happens (Hive creates multiple counters per operator), how to raise mapreduce.job.counters.max safely, and how to estimate how many operators your query uses with EXPLAIN so you can tune the setting without guessing.

Symptom

A Hive job fails with an error like: 

Ended Job = job_xxxxxx with exception
'org.apache.hadoop.mapreduce.counters.LimitExceededException
 (Too many counters: 201 max=200)'
FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.MapRedTask
Intercepting System.exit(1)

The query might be complex, but otherwise looks syntactically fine. The problem is not your SQL, it’s the number of counters that MapReduce is willing to track.

Why this happens: operators and counters

Hive uses counters to track statistics for built-in operators (see the Hive LanguageManual UDF / operators documentation). For each operator, Hive can create multiple counters; historically this was around four counters per operator.

On top of that, a MapReduce job has its own counters for:

  • Input / output bytes and records
  • Tables, partitions, files
  • Custom job- and framework-level metrics

The default maximum number of counters in classic MapReduce was relatively low (for example, 200), so a sufficiently complex query with many operators could exceed this limit and trigger a LimitExceededException.

The fix: raise mapreduce.job.counters.max

To avoid this exception, increase the maximum number of counters allowed per job by setting: 

<property>
  <name>mapreduce.job.counters.max</name>
  <value>1120</value>
</property>

Add this to mapreduce-site.xml (or the equivalent config for your distribution) and restart the relevant services so Hive and MapReduce pick up the new setting.

A value slightly above 1000 is often enough for complex Hive queries while still keeping memory overhead manageable. You can adjust this based on your query patterns and any additional counters used by your environment.

Estimating how many counters you need

The number of counters required is roughly proportional to the number of operators used in the query. A simple way to get a sense of this is:

  1. Run EXPLAIN EXTENDED on your query to see the full operator tree.
  2. Search for “operator” lines in the explain output and count them.

Example workflow:

hive> EXPLAIN EXTENDED <your_query> > explain.out

grep -ri "operator" explain.out | wc -l

This gives you a rough count of operators. Multiply by the number of counters per operator (historically ~4) and add some headroom for framework counters to decide how high you should set mapreduce.job.counters.max.

Impact on other MapReduce jobs

Raising mapreduce.job.counters.max:

  • Prevents Hive jobs from failing as they approach the previous counter limit.
  • Does not break other MapReduce jobs, but they can now allocate more counters if needed.
  • Has a small memory and bookkeeping cost per additional counter; keep the value reasonable and not arbitrarily huge.

If you see the “Too many counters” error, adjusting this single setting is usually enough to unblock Hive while keeping your MapReduce cluster stable.

Related guides:

If you need help with distributed systems, backend engineering, or data platforms, check my Services.

Labels:

Most read articles

Building a Model-Agnostic Multi-Agent System with OpenClaw

By
Over one week we rebuilt our AI stack around OpenClaw’s multi-agent architecture to avoid provider lock-in and stop wasting premium tokens. By aligning models to tasks, diversifying fallbacks across providers, enforcing minimal tool access, and switching to memory-first workflows with ephemeral sessions, we reduced token usage per task by about 70% and cut our monthly bill by 77% while improving operational resilience. How We Achieved 77% Cost Reduction and Provider Independence Over the past week, we rebuilt our AI infrastructure around OpenClaw’s multi-agent architecture. The result was a 77% cost reduction , provider independence , and a delegation system that routes work to the most cost-effective model for each job. Below is the technical journey of optimizing a 7-agent squad with OpenClaw. The Challenge: Model Provider Lock-In We started with a simple problem: our entire squad defaulted to a single model provider. This created three issues: Cost inefficiency beca...
Read more

Why Is Customer Obsession Disappearing?

By
Many companies trade real customer-obsession for automated, low-empathy support. Through examples from Coinbase, PayPal, GO Telecommunications and AT&T, this article shows how reliance on AI chatbots, outsourced call centers, and KPI-driven workflows erodes trust, NPS and customer retention. It argues that human-centric support—treating support as strategic investment instead of cost—is still a core growth engine in competitive markets. It's wild that even with all the cool tech we've got these days, like AI solving complex equations and doing business across time zones in a flash, so many companies are still struggling with the basics: taking care of their customers. The drama around Coinbase's customer support is a prime example of even tech giants messing up. And it's not just Coinbase — it's a big-picture issue for the whole industry. At some point, the idea of "customer obsession" got replaced with "customer automation," and no...
Read more

What are the performance implications of cross-platform execution within Wayang?

By
Apache Wayang ® enables cross-platform execution across multiple data processing platforms such as Spark, Flink, Java Streams, PostgreSQL or GraphChi. This capability fundamentally changes the performance behavior of distributed data pipelines. Wayang reduces manual data movement by selecting where each operator should run, but crossing platform boundaries still introduces serialization cost, shifts in locality, different memory strategies and new tuning constraints. Understanding these dynamics is essential before adopting Wayang for multi-platform pipelines at scale. Apache Wayang is a cross-platform data processing framework that lets developers run a single logical pipeline across engines such as Apache Spark, Apache Flink or a native Java backend. It provides an abstraction layer and a cost-based optimizer that selects the execution platform for each operator. This flexibility introduces new performance variables that do not exist in single-engine systems. Engine boundaries ...
Read more