The master server is Informix, version varies from 9.40 to the latest, database is unlogged by design that can't be changed. Slave server is the latest PostgreSQL. Master and slave are separate machines, network latency is unpredictable. Master schema is statically defined, well known and does not change, so it's only the data that needs to be replicated. In the master, there are three types of tables:
Numeric data tables, usually one date column, one time column and 15-300 int columns keyed by 2-3 primary keys. The data is never changed, only added once in a set interval (15, 30, or 60 minutes) and deleted when the retention point is reached. Replication data set can be up to 80,000 rows but usually is in the range of hundreds. This data needs to be replicated one way, master to slave. There is about 30 tables of this type and they need to be replicated all at once and as fast as possible, typically in under one minute after new interval set has been committed to the master.
Mixed data tables, with date, time, int, and string types, 30-100 columns, again 2-3 primary keys. This data is also never changed, added continuously and is deleted when the retention point is reached. The data set is up to 100,000 rows per hour. One way replication is needed, master to slave. There are a few tables like that, less than 5 usually.
Mixed data tables, with int and string types, less than 10 columns, 2-3 primary keys. The data largely stays intact, with occasional additions, edits or deletions. The usual replication set size is unpredictable, but probably will be in low hundreds of rows. This data needs to be replicated both ways, as fast as possible. There are a few tables of this type, and they need to be synched independently.
I've been looking for an existing tool that could do what I need, but it looks like there is none that is open source. I'm probably going to write one for my needs, and I'm looking for advice from DB gurus on how to approach this task.
In my estimate, there's probably no single algorithm that would cover all the use cases so I may be in fact looking for two or three algorithms. Here's what I found so far:
Fire trigger on master changes, record row OIDs (does Informix have them?) to temp table, dump the changed rows to a file, transfer it and load up. Question: how to buffer the trigger? The master DB is unlogged (no transactions), so trigger will fire upon each INSERT. Additional strain on the master, not good.
Add a cron job on the slave that will pull latest date/time keys from the master, and if the data is newer, pull it. Problem: although the update interval is defined, in reality it's based on the data source clock (not master DB clock) which is guaranteed to vary from slave server clock. More of it, there can be several data sources, each with varying clocks, and the data needs to be replicated ASAP. The only way here that I see is to constantly poll the master from the slave, hoping that by the time the poll comes in, the data is all committed (no transactions, remember?). Kludgy, slow, not good.
Add Informix as foreign data wrapper in the Postgres and run queries directly instead of bothering with replication. Pros: simplicity. Cons: Informix connector seems to be in alpha stage, and the whole approach is an unknown factor at best.
I've been researching this topic for some time, and it seems that the core of the problem is the lack of transactions on the master side. If the master DB was logged, it would be much easier to replicate it, but without transactions the task suddenly becomes much more complicated. For one, how do I ensure that there are no dupes? Another one, how to avoid update loops in type 3 tables? Considering all that, how to make replication as fast-reacting as possible? I mean the delay between data update and sync start here, data transfer is another topic altogether.
Any input is appreciated.
If you can't change the master in any significant way you are going to have a heck of a time with any sort of replication. Your basic problem is that you have no real way to handle replicating changes in real time without tracking which changes have been replicated, and if you can't change the master, you can't add that. So the short answer is that replication is not a solution which can work for you. Given some of Informix's other features I would think twice about going about this as continuous replication.
This leads to other approaches. The big unknown factors are that networks may not be reliable enough to just link the databases. This could lead to transactions hanging while waiting for data off a high latency connection to all kinds of other problems. You might be able to get this to work with an odbc fdw and an informix provider or with DBI-Link and DBD::Informix, but this strikes me as a problem in your current environment. You could use these in a cron job to populate a second PostgreSQL server closer to your own location periodically, however and so I would not write the approach entirely off.
One way or another it seems to me you need to get a copy of the data to your PostgreSQL server. You may want to do an ETL job to import the data periodically. You may want to use a secondary postgresql server and FDW's or DBI-Link to pull in the data. But this is not likely to be real-time, it is not likely to be continuous.
The tl;dr is that your environment isn't really set up to do this. For my money I would recommend an ETL approach and accept that your slave will not be in sync with the master.
Related
On my primary I ran a VACUUM then an ANALYZE on all databases, then when I check pg_stat_user_tables, the last_analyze column shows a current timestamp which is great.
When I check my replication instance, there are no values in the last_analyze column. I was assuming this timestamp would also eventually populate? Is this known behaviour?
The reason I ask is that after that VACUUM/ANALYZE on the primary, I'm running into some extremely slow queries on the replication instance. I ran an EXPLAIN plan prior to the VACUUM/ANALYZE on a query and it ran in 5 seconds... now it's taking 65 seconds. The EXPLAIN shows it's not using a lot of indexes that it should be.
PostgreSQL has two different stats systems. One records data about the distribution of values in the columns, this is transactional. It propagates to the replica via the WAL.
The other system records data about turn over on the tables and data on when the last vac/an was done. This system is used to determine when to schedule new vac/an (to prevent the first system from getting too out of date). This one is not transactional, and does not propagate to the replica.
So the replica has the latest column value distribution statistics (as soon as the WAL replays, anyway), but it doesn't know how recent they are.
If you have scaled SQL server with one DB for writes and multiple DBs for reads. Wouldn't there be a delay for data to be replicated from the write DB to the to other read databases? In which case isn't the data inconsistent?
So where would a scaled relational DB fall in the CAP theorem?
Update:
In relational DBs consistency means there wont be partial updates. For example if someone transfers money from one account to another and the whole thing is a part of one transaction, it wont happen that you take money out of one account but doesn't show up in another account.
In CAP theorem consistence means all the components see the same data. That consistency is different from consistency in ACID.
From what I know, relational DBs like SQL server are supposed to be CA (consistent and available). This would make sense if there is just one database. Because everyone would see the same data. But what if the SQL server is scaled with multiple databases? In that case would all databases still see the same data? If not, would it be consistent (in CAP theorem)?
My feeling is a scaled relational DB is AP (Available and partition tolerant) and not CA (Consistent and available).
I've read different definitions of consistency in regards to the CAP theorem.
Some definitions of consistency say that once some data is persisted in a system, all reads will read the most recently written data. In this definition, a replicated database (you call this "scaled" but I wouldn't use that term) has a risk of returning inconsistent data, if the replication is asynchronous.
To mitigate this risk, some systems make sure replication is synchronous, or as close to synchronous as they can implement. Galera, for example, sends transaction write sets to its replicas synchronously. If you try to read from the replica, and it detects that there are write sets pending but not yet applied, it can block your read until it has caught up with the pending write sets (this behavior is configurable). So you'll never read data that is out of date.
The cost of maintaining perfectly consistent reads over distributed systems in this manner is usually more expensive than users want. It will become a performance bottleneck in a system that has a high rate of updates. So for practical reasons, most projects accept that "replication lag" is a necessary compromise.
Other definitions of consistency are closer to atomicity, i.e. transactions will not be persisted in a partially-complete state. So all constraints will be satisfied when you read the data, whether you read the data before or after the transaction is applied. In this definition, it's quite easy to imagine the replica database instance remaining consistent, if it applies updates using the same transaction semantics used on the master. If you read data from the replica, you might read data that hasn't yet had the latest updates applied, but it will never be in an inconsistent state with respect to constraints.
There is nothing called a scaled RDBMS. We do have "RDBMS Clusters with shared storage": here can keep on adding nodes to achieve high availability of RDBMS.
In other words:
If you meant a "Distributed RDBMS" by mentioning "Scaled RDBMS" - it doesn't exist. You can have RDBMS on only one node. If you add another node, then that will be "another" RDBMS and it would NOT coalesce with the first one giving you a single view(unlike a typical NoSQL Database). Although, you can happily keep on adding storage nodes behind the RDBMS.
Is it possible to use a table in cassandra as a queue, I don't think the strategy I use in mysql works, ie given this table:
create table message_queue(id integer, message varchar(4000), retries int, sending boolean);
We have a transaction that marks the row as "sending", tries to send, and then either deletes the row, or increments the retries count. The transaction ensures that only one server will be attempting to process an item from the message_queue at any one time.
There is an article on datastax that describes the pitfalls and how to get around it, however Im not sure what the impact of having lots of tombstones lying around is, how long do they stay around for?
Don't do this. Cassandra is a terrible choice as a queue backend unless you are very, very careful. You can read more of the reasons in Jonathan Ellis blog post "Cassandra anti-patterns: Queues and queue-like datasets" (which might be the post you're alluding to). MySQL is also not a great choice for backing a queue, us a real queue product like RabbitMQ, it's great and very easy to use.
The problem with using Cassandra as the storage for a queue is this: every time you delete a message you write a tombstone for that message. Every time you query for the next message Cassandra will have to trawl through those tombstones and deleted messages and try to determine the few that have not been deleted. With any kind of throughput the number of read values versus the number of actual live messages will be hundreds of thousands to one.
Tuning GC grace and other parameters will not help, because that only applies to how long tombstones will hang around after a compaction, and even if you dedicated the CPUs to only run compactions you would still have dead to live rations of tens of thousands or more. And even with a GC grace of zero tombstones will hang around after compactions in some cases.
There are ways to mitigate these effects, and they are outlined in Jonathan's post, but here's a summary (and I don't write this to encourage you to use Cassandra as a queue backend, but because it explains a bit more about Cassandra works, and should help you understand why it's a bad fit for the problem):
To avoid the tombstone problem you cannot keep using the same queue, because it will fill upp with tombstones quicker than compactions can get rid of them and your performance will run straight into a brick wall. If you add a column to the primary key that is deterministic and depends on time you can avoid some of the performance problems, since fewer tombstones have time to build up and Cassandra will be able to completely remove old rows and all their tombstones.
Using a single row per queue also creates a hotspot. A single node will have to handle that queue, and the rest of the nodes will be idle. You might have lots of queues, but chances are that one of them will see much more traffic than the others and that means you get a hotspot. Shard the queues over multiple nodes by adding a second column to the primary key. It can be a hash of the message (for example crc32(message) % 60 would create 60 shards, don't use a too small number). When you want to find the next message you read from all of the shards and pick one of the results, ignoring the others. Ideally you find a way to combine this with something that depends on time, so that you fix that problem too while you're at it.
If you sort your messages after time of arrival (for example with TIMEUUID clustering key) and can somehow keep track of the newest messages that has been delivered, you can do a query to find all messages after that message. That would mean less thrawling through tombstones for Cassandra, but it is no panacea.
Then there's the issue of acknowledgements. I'm not sure if they matter to you, but it looks like you have some kind of locking mechanism in your schema (I'm thinking of the retries and sending columns). This will not work. Until Cassandra 2.0 and it's compare-and-swap features there is no way to make that work correctly. To implement a lock you need to read the value of the column, check if it's not locked, then write that it should now be locked. Even with consistency level ALL another application node can do the same operations at the same time, and both end up thinking that they locked the message. With CAS in Cassandra 2.0 it will be possible to do atomically, but at the cost of performance.
There are a couple of more answers here on StackOverflow about Cassandra and queues, read them (start with this: Table with heavy writes and some reads in Cassandra. Primary key searches taking 30 seconds.
The grace period can be defined. Per default it is 10 days:
gc_grace_seconds¶
(Default: 864000 [10 days]) Specifies the time to wait before garbage
collecting tombstones (deletion markers). The default value allows a
great deal of time for consistency to be achieved prior to deletion.
In many deployments this interval can be reduced, and in a single-node
cluster it can be safely set to zero. When using CLI, use gc_grace
instead of gc_grace_seconds.
Taken from the
documentation
On a different note, I do not think that implementing a queue pattern in Cassandra is very useful. To prevent your worker to process one entry twice, you need to enforce "ALL" read consistency, which defeats the purpose of distributed database systems.
I highly recommend looking at specialized systems like messaging systems which support the queue pattern natively. Take a look at RabbitMQ for instance. You will be up and running in no time.
Theo's answer about not using Cassandra for queues is spot on.
Just wanted to add that we have been using Redis sorted sets for our queues and it has been working pretty well. Some of our queues have tens of millions of elements and are accessed hundreds of times per second.
I am working on a project which has some important data in it. This means we cannot to lose any of it if the light or server goes down. We are using MongoDB for the database. I'd like to be sure that my data is in the database after the insert and rollback the whole batch if one element was not inserted. I know it is the philosophy behind Mongo that we do not need transactions but how can I make sure that my data is really safely stored after insert rather than sent to some "black hole".
Should I make a search?
Should I use some specific mongoDB commands?
Should I use sharding even if one server is enough for satisfying
the speed and by the way it doesn't guarantee anything if the light
goes down?
What is the best solution?
Your best bet is to use Write Concerns - these allow you to tell MongoDB how important a piece of data is. The quickest Write Concern is also the least safe - the data is not flushed to disk until the next scheduled flush. The safest will confirm that the data has been written to disk on a number of machines before returning.
The write concern you are looking for is FSYNC_SAFE (at least that is what it is called from the point of view of the Java driver) or REPLICAS_SAFE which confirms that your data has been replicated.
Bear in mind that MongoDB does not have transactions in the traditional sense - your rollback will have to be rolled by hand as you can't tell the Mongo database to do this for you.
The other thing you need to do is either use the relatively new --journal option (which uses a Write Ahead Log), or use replica sets to share your data across many machines in order to maximise data integrity in the event of a crash/power loss.
Sharding is not so much a protection against hardware failure as a method for sharing the load when dealing with particularly large datasets - sharding shouldn't be confused with replica sets which is a way of writing data to more than one disk on more than one machine.
Therefore, if your data is valuable enough, you should definitely be using replica sets, perhaps even siting slaves in other data centres/availability zones/racks/etc in order to provide the resilience you require.
There is/will be (can't remember offhand whether this has been implemented yet) a way to specify the priority of individual nodes in a replica set such that if the master goes down the new master that is elected is one in the same data centre if such a machine is available (ie to stop a slave on the other side of the country from becoming master unless it really is the only other option).
I received a really nice answer from a person called GVP on google groups. I will quote it(basically it adds up to Rich's answer):
I'd like to be sure that my data is in the database after the
insert and rollback the whole batch if one element was not inserted.
This is a complex topic and there are several trade-offs you have to
consider here.
Should I use sharding?
Sharding is for scaling writes. For data safety, you want to look a
replica sets.
Should I use some specific mongoDB commands?
First thing to consider is "safe" mode or "getLastError()" as
indicated by Andreas. If you issue a "safe" write, you know that the
database has received the insert and applied the write. However,
MongoDB only flushes to disk every 60 seconds, so the server can fail
without the data on disk.
Second thing to consider is "journaling"
(v1.8+). With journaling turned on, data is flushed to the journal
every 100ms. So you have a smaller window of time before failure. The
drivers have an "fsync" option (check that name) that goes one step
further than "safe", it waits for acknowledgement that the data has
be flushed to the disk (i.e. the journal file). However, this only
covers one server. What happens if the hard drive on the server just
dies? Well you need a second copy.
Third thing to consider is
replication. The drivers support a "W" parameter that says "replicate
this data to N nodes" before returning. If the write does not reach
"N" nodes before a certain timeout, then the write fails (exception
is thrown). However, you have to configure "W" correctly based on the
number of nodes in your replica set. Again, because a hard drive
could fail, even with journaling, you'll want to look at replication.
Then there's replication across data centers which is too long to get
into here. The last thing to consider is your requirement to "roll
back". From my understanding, MongoDB does not have this "roll back"
capacity. If you're doing a batch insert the best you'll get is an
indication of which elements failed.
Here's a link to the PHP driver on this one: http://it.php.net/manual/en/mongocollection.batchinsert.php You'll have to check the details on replication and the W parameter. I believe the same limitations apply here.
My web app uses ADO.NET against SQL Server 2008. Database writes happen against a primary (publisher) database, but reads are load balanced across the primary and a secondary (subscriber) database. We use SQL Server's built-in transactional replication to keep the secondary up-to-date. Most of the time, the couple of seconds of latency is not a problem.
However, I do have a case where I'd like to block until the transaction is committed at the secondary site. Blocking for a few seconds is OK, but returning a stale page to the user is not. Is there any way in ADO.NET or TSQL to specify that I want to wait for the replication to complete? Or can I, from the publisher, check the replication status of the transaction without manually connecting to the secondary server.
[edit]
99.9% of the time, The data in the subscriber is "fresh enough". But there is one operation that invalidates it. I can't read from the publisher every time on the off chance that it's become invalid. If I can't solve this problem under transactional replication, can you suggest an alternate architecture?
There's no such solution for SQL Server, but here's how I've worked around it in other environments.
Use three separate connection strings in your application, and choose the right one based on the needs of your query:
Realtime - Points directly at the one master server. All writes go to this connection string, and only the most mission-critical reads go here.
Near-Realtime - Points at a load balanced pool of subscribers. No writes go here, only reads. Used for the vast majority of OLTP reads.
Delayed Reporting - In your environment right now, it's going to point to the same load-balanced pool of subscribers, but down the road you can use a technology like log shipping to have a pool of servers 8-24 hours behind. These scale out really well, but the data's far behind. It's great for reporting, search, long-term history, and other non-realtime needs.
If you design your app to use those 3 connection strings from the start, scaling is a lot easier, especially in the case you're experiencing.
You are describing a synchronous mirroring situation. Replication cannot, by definition, support your requirement. Replication must wait for a transaction to commit before reading it from the log and delivering it to the distributor and from there to the subscriber, which means replication by definition has a window of opportunity for data to be out of sync.
If you have a requirement an operation to read the authorithative copy of the data, then you should make that decission in the client and ensure you read from the publisher in that case.
While you can, in threory, validate wether a certain transaction was distributed to the subscriber or not, you should not base your design on it. Transactional replication makes no latency guarantee, by design, so you cannot rely on a 'perfect day' operation mode.