Skip to content
This repository has been archived by the owner on Aug 23, 2021. It is now read-only.

SEATS: Out Of memory error #330

Open
stevalla opened this issue Jul 2, 2020 · 0 comments
Open

SEATS: Out Of memory error #330

stevalla opened this issue Jul 2, 2020 · 0 comments

Comments

@stevalla
Copy link

stevalla commented Jul 2, 2020

Using the SEATS benchmark against Postgres I found some OOM problems with some specific transaction types. In particular with two workloads: the first using 100% "New reservation" transactions, the second using 100% "Delete reservation" transactions.

The workload configuration is the following:

<?xml version="1.0"?>
<parameters>
    <!-- Scalefactor in SEATS is *1000 the number of customers (200 = 48 GB)-->
    <scalefactor>53</scalefactor>

    <terminals>50</terminals>
    <works>
        <work>
            <time>1200</time>
            <rate>unlimited</rate>
            <weights>
                100, 0, 0, 0, 0, 0
            </weights>
        </work>
    </works>

    <transactiontypes>
        <transactiontype>
                <name>DeleteReservation</name>
        </transactiontype>
        <transactiontype>
                <name>FindFlights</name>
        </transactiontype>
        <transactiontype>
                <name>FindOpenSeats</name>
        </transactiontype>
        <transactiontype>
                <name>NewReservation</name>
        </transactiontype>
        <transactiontype>
                <name>UpdateCustomer</name>
        </transactiontype>
        <transactiontype>
                <name>UpdateReservation</name>
        </transactiontype>
    </transactiontypes>
</parameters>

The machine where the generator is running has 16 GB of rams and 4 CPUs.

When the test starts, OLTP starts to fill the memory until it is totally saturated and it crashes with an OOM error. Furthermore, during the test there is no activity on the Postgres machine.

It shouldn't be a connection issue since I change the transaction mix with the same configuration file and it works. Thus, I think it is something related to the two types of transactions listed above.

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant