Skip to main content

Walking down the Memory Lane!!!

This post is going to be an account of  few trouble-shootings I did recently to combat various I/O sluggishness.
  1. Slow system during problems with backup
    We have a NFS mount where we push backups of our database daily. Due to some update to the NFS infra, we started seeing throughput of NFS server drastically affected. During this time we saw general sluggishness in the system during backups. Even ssh logins appeared slower. Some boxes had to be rebooted due to this sluggishness as they were too slow to operate on them. First question we wanted to answer, does NFS keep writing if the server is slow? The slow server applied back pressure by sending small advertised window(TCP) to clients. So clients can't push huge writes if server is affected. Client writes to its page cache. The data from page cache is pushed to server when there is a memory pressure or file close is called. If server is slow, client can easily reach upto dirty_background_ratio set for page cache in sysctl. This dictates how much dirty pages cache can have maximum. NFS has its own implementation of bdi(backing device information) This rolling congestion control applies back pressure to linux user space process which calls write syscall and the user space process stops writing. During this time, kswapd becomes aggressive to keep the dirty pages under dirty_background_ratio and starts writing any data from any filesystem as fast as it can write to disk. kswapd will be aggressive for time NFS is being slow or dirty pages comes below the configured sysctl value. This will impact all apps which needs logging or file write operation as all writes should hit disk. BDI does support controlling how much dirty ratio each device and virtual devices like NFS can take. Looks like treading down that path should avoid such problems
  2. Page cache approaching zero size despite tons of memory
    While debugging latency issue, team identified page cache stayed closer to 0G for hours despite usual I/O on system. The system had tons of memory close to 10G. On further debugging I noticed kswapd is in D state on affected systems. kswapd in D state quickly led to this link This link said this could be due to any kernel module requesting contiguous memory pages. Contiguous physical memory is something that is not required in today's world post IO_MMU. But if some driver is written with still old practices, they can request contiguous pages. Right now if contiguous pages are not available kswapd is called, kswapd continuously flushes pages to disk till it gets requested number of contiguous memory pages available. This dumb approach of kswapd(captured here lets it flush as much as it can till it gets contiguous memory. When we triggered online memory compact by setting vm.compact_memory in sysctl, we started noticing page cache to behave normally and kswapd jumped back to S state. We have zeroed down the bug to a driver version of one our hardware vendors using slabtop.
  3. High latency when deleting huge file
    We observed spike in latency to our clients when our service actually deletes databases which were no longer needed. Deletion of database actually calls unlink syscall. Though initial assumption assumed deletes caused higher I/O and hence latency, its not how file systems work. Filesystems usually mark blocks free but don't zero them on disk. This is very well the reason why so many disk recovery tools exist. For the lifetime of unlink, we figured out the DB server has a global mutex blocking opening or closing of new tables. This translates to latency. Unlink can take time in filesystems since it has to mark all blocks of file as free despite ext4 boasts significant improvement in performance compared to ext3 due to extents. Based on my understanding only the thread which calls unlink should be affected due to the deletion but global lock kind of spreads the issue across the app.


Popular posts from this blog

How we have systematically improved the roads our packets travel to help data imports and exports flourish

This blog post is an account of how we have toiled over the years to improve the throughput of our interDC tunnels. I joined this company around 2012. We were scaling aggressively then. We quickly expanded to 4 DCs with a mixture of AWS and colocation. Our primary DC is connected to all these new DCs via IPSEC tunnels established from SRX. The SRX model we had, had an IPSEC throughput of 350Mbps. Around December 2015 we saturated the SRX. Buying SRX was an option on the table. Buying one with 2Gbps throughput would have cut the story short. The tech team didn't see it happening. I don't have an answer to the question, "Is it worth spending time in solving a problem if a solution is already available out of box?" This project helped us in improving our critical thinking and in experiencing the theoretical network fundamentals on live traffic, but also caused us quite a bit of fatigue due to management overhead. Cutting short the philosophy, lets jump to the story.

LXC and Host Crashes

 We had set up a bunch of lxc containers on two servers each with 16 core CPUs and 64 GB RAM(for reliability and loadbalancing). Both the servers are on same vlan. The servers need to have atleast one of their network interface in promiscuous mode so that it forwards all packets on vlan to the bridge( ) which takes care of the routing to containers. If the packets are not addressed to the containers, the bridge drops the packet. Having this setup, we moved all our platform maintenance services to these containers. They are fault tolerant as we used two host machines where each host machine has a replica of the containers on the other. The probability to crash for both the servers at the same time due to some hardware/software failure is less. But to my surprise both the servers are crashing exactly the same time with a mean life time 20 days. We had to wake up late nights(early mornings) to fix stuffs that gone down The

The server, me and the conversation

We were moving a project from AWS to our co-located DC. We have setup KVMs scheduled by Cloudstack for each of the component in the architecture. The KVMs used local storage. The VMs are provisioned with more than required resources because we have the opinion that in our DC scaling during peak load and then downscaling doesn't offer much benefits financially as we are anyways paying for the hardware in advance and its also powered on. Its going to be idle if not used. Now we found something interesting our latency in co-located DC was 2 times more than in AWS. The time for first byte at our load balancer in aws was 60ms average and at our DC was 112ms. We started our debugging mission, Mission Conquer-AWS. All the servers are newer Dell hardwares. So the initially intuition was virtualisation is causing the issue. Conversation with the Hypervisor We started with CPU optimisation, we started using the host-passthrough mode of CPU in libvirt so VMs dont see QEMU emulated CPUs,