Folks, If you have a replication factor set to 1 (aka. backup) then a loss of a primary node won't affect the availability of your cluster as long as the backup will be serving the requests.
- Denis On Wed, May 22, 2019 at 9:03 PM Coleman, JohnSteven (Agoda) < [email protected]> wrote: > Hi, > > Ignite can go so far in meeting your requirements for durable storage and > retrieval, but a primary node failure is probably going to be terminal > because unless you can boot a machine in less than 5 seconds you can forget > about reloading ignite anyway, you simply won't be able to restore your TCP > connection in time. (Unless of course you have some radical kind of > hardware with OS in bubble memory or similar, but even then?) > > So you need to use something like a p2p protocol such as bit torrent where > the download is broken into chunks and delegated to multiple nodes each > holding some or all of the blobs data. The client has to be able to recover > as well, either by round robin or by implementing a p2p protocol - you > still can't depend on a single fallible point of entry to your service > because of the problem above. > > If you have such a p2p protocol I'm not sure Ignite adds anything much of > value? > > John > > > -----Original Message----- > From: steven <[email protected]> > Sent: Thursday, May 23, 2019 2:16 AM > To: [email protected] > Subject: Is Apache Ignite appropriate for my use case? > > Email received from outside the company. If in doubt don't click links nor > open attachments! > ________________________________ > > Hi, I need to manage a large fleet of servers (around 100k machines). Each > node contains a subset of the data (basically binary blobs) in memory. The > data store is a separate database. > > I am trying to provide a binary blob lookup service (in greatly simplified > terms). For each incoming request (which contains a binary string), every > node should perform matching of the string against its subset of the data. > When the first matching binary blob is found, that result is returned, and > all other nodes should stop searching. If no matches are found, NOT_FOUND > should be returned. Requests must be handled within 5 seconds. > > If any node fails, it must be revived with the same subset of data that it > had before. > > One challenge is how to handle node failure in the middle of a request. It > is unlikely that a node will be revived quickly enough to respond within 5 > seconds. Most likely there should be standby nodes that will retrieve the > failing node's subset of data from the data store and perform matching upon > being notified of node failure. > > Is Apache Ignite appropriate for this use case? > > Thanks, > Steven > > > > -- > Sent from: http://apache-ignite-users.70518.x6.nabble.com/ > > ________________________________ > This message is confidential and is for the sole use of the intended > recipient(s). It may also be privileged or otherwise protected by copyright > or other legal rules. If you have received it by mistake please let us know > by reply email and delete it from your system. It is prohibited to copy > this message or disclose its content to anyone. Any confidentiality or > privilege is not waived or lost by any mistaken delivery or unauthorized > disclosure of the message. All messages sent to and from Agoda may be > monitored to ensure compliance with company policies, to protect the > company's interests and to remove potential malware. Electronic messages > may be intercepted, amended, lost or deleted, or contain viruses. >
