Skip to main content

What a rocky start to labor day weekend

Woke up by earthquake at 7:00 AM in morning and then couldn't get to sleep. I took a bath, made my tea and started checking emails and saw that after last night deployment three storage node out of 100s of nodes were running into Full GC. What was special about the 3 nodes was that each one was in a different Data centre but it was named same app02.  This got me curious I asked the node to be taken out of rotation and take a heap dump.  Yesterday night a new release has happened and I had upgraded spymemcached library version as new relic now natively supports instrumentation on it so it was a suspect. And the hunch was a bullseye, the heap dump clearly showed it taking 1.3G and full GCs were taking 6 sec but not claiming anything.



I have a quartz job in each jvm that takes a thread dump every 5 minutes and saves last 300 of them, checking few of them quickly showed a common thread among all 3 data centres. It seems there was a long running job that was trying to replicate pending storage objects as per replication policy and it had 50M or so backlog to catchup.  It seems newrelic was holding on to all that data for the thread and not just collecting the aggregate stats, the job was running for hours causing memory to increase with each replicated file.

Now I had two options:-
  1. Rollback the library
  2. Suppress newrelic instrumentation
I thought of going option#2 as it wont require any new build, just run puppet and restart all nodes.  Boy I was wrong , newrelic has negligible documentation for the yml file.  I filed a support ticket as they have no phone support even for emergencies even though we pay thousands of dollars a month. After IPO their support team quality has also reduced, the support told me hey go and disable the class, i am like its natively supported by you so give me some hints how to do it. She told me that a OOM ticket is filed 2 versions ago so I can wait for it to be fixed, wtf. I waited 20 min for her to tell me how to disable it and no reply,  It seems even she didnt knew how to do it so she was buying time letting me hanging dry on a cloth line. I started mining the support forum and google and I tried 5-10 settings with our devops engineer and newrelic agent is stupid that it would take any wrong setting too and show it in the UI. I ended up with a settings nightmare


Finally I tried one setting which was weird but worked, I had updated library to spymemcached-2.12.1 but the setting to turn it off in newrelic was

  class_transformer:
    com.newrelic.instrumentation.spymemcached-2.12.0:
      enabled: false

notice the setting is called 2.12.0 even though my library is 2.12.1, wtf.  It was late in night for the devops engineer and I was almost 5 hours on the desk, we just patched those 3 special nodes and called it a day. The other nodes would be patched by puppet before Monday using automation. 

Today I also complete 7  years at the company :).


Comments

Popular posts from this blog

Killing a particular Tomcat thread

Update: This JSP does not work on a thread that is inside some native code.  On many occasions I had a thread stuck in JNI code and it wont work. Also in some cases thread.stop can cause jvm to hang. According to javadocs " This method is inherently unsafe. Stopping a thread with Thread.stop causes it to unlock all of the monitors that it has locked". I have used it only in some rare occasions where I wanted to avoid a system shutdown and in some cases we ended up doing system shutdown as jvm was hung so I had a 70-80% success with it.   -------------------------------------------------------------------------------------------------------------------------- We had an interesting requirement. A tomcat thread that was spawned from an ExecutorService ThreadPool had gone Rogue and was causing lots of disk churning issues. We cant bring down the production server as that would involve downtime. Killing this thread was harmless but how to kill it, t

Adding Jitter to cache layer

Thundering herd is an issue common to webapp that rely on heavy caching where if lots of items expire at the same time due to a server restart or temporal event, then suddenly lots of calls will go to database at same time. This can even bring down the database in extreme cases. I wont go into much detail but the app need to do two things solve this issue. 1) Add consistent hashing to cache layer : This way when a memcache server is added/removed from the pool, entire cache is not invalidated.  We use memcahe from both python and Java layer and I still have to find a consistent caching solution that is portable across both languages. hash_ring and spymemcached both use different points for server so need to read/test more. 2) Add a jitter to cache or randomise the expiry time: We expire long term cache  records every 8 hours after that key was added and short term cache expiry is 2 hours. As our customers usually comes to work in morning and access the cloud file server it can happe

Preparing for an interview after being employed 11 years at a startup

I would say I didn't prepared a hell lot but  I did 2 hours in night every day and every weekend around 8 hours for 2-3 months. I did 20-30 leetcode medium problems from this list https://leetcode.com/explore/interview/card/top-interview-questions-medium/.  I watched the first 12 videos of Lecture Videos | Introduction to Algorithms | Electrical Engineering and Computer Science | MIT OpenCourseWare I did this course https://www.educative.io/courses/grokking-the-system-design-interview I researched on topics from https://www.educative.io/courses/java-multithreading-for-senior-engineering-interviews and leetcode had around 10 multithreading questions so I did those I watched some 10-20 videos from this channel https://www.youtube.com/channel/UCn1XnDWhsLS5URXTi5wtFTA